机器学习学术速递[1.21]

点击阅读原文访问arxivdaily.com，涵盖CS|物理|数学|经济|统计|金融|生物|电气领域，更有搜索、收藏等功能！

cs.LG 方向，今日共计383篇

大模型相关(53篇)

【1】InT: Self-Proposed Interventions Enable Credit Assignment in LLM Reasoning
标题：InT：自我提议的干预使LLM推理中的信用分配成为可能
链接：https://arxiv.org/abs/2601.14209

作者：Matthew Y. R. Yang,Hao Bai,Ian Wu,Gene Yang,Amrith Setlur,Aviral Kumar
摘要：Outcome-reward reinforcement learning (RL) has proven effective at improving the reasoning capabilities of large language models (LLMs). However, standard RL assigns credit only at the level of the final answer, penalizing entire reasoning traces when the outcome is incorrect and uniformly reinforcing all steps when it is correct. As a result, correct intermediate steps may be discouraged in failed traces, while spurious steps may be reinforced in successful ones. We refer to this failure mode as the problem of credit assignment. While a natural remedy is to train a process reward model, accurately optimizing such models to identify corrective reasoning steps remains challenging. We introduce Intervention Training (InT), a training paradigm in which the model performs fine-grained credit assignment on its own reasoning traces by proposing short, targeted corrections that steer trajectories toward higher reward. Using reference solutions commonly available in mathematical reasoning datasets and exploiting the fact that verifying a model-generated solution is easier than generating a correct one from scratch, the model identifies the first error in its reasoning and proposes a single-step intervention to redirect the trajectory toward the correct solution. We then apply supervised fine-tuning (SFT) to the on-policy rollout up to the point of error concatenated with the intervention, localizing error to the specific step that caused failure. We show that the resulting model serves as a far better initialization for RL training. After running InT and subsequent fine-tuning with RL, we improve accuracy by nearly 14% over a 4B-parameter base model on IMO-AnswerBench, outperforming larger open-source models such as gpt-oss-20b.

【2】Lost in the Prompt Order: Revealing the Limitations of Causal Attention in Language Models
标题：迷失在提示顺序中：揭示语言模型中因果注意的局限性
链接：https://arxiv.org/abs/2601.14152

作者：Hyunjong Ok,Jaeho Lee
备注：preprint
摘要：Large language models exhibit surprising sensitivity to the structure of the prompt, but the mechanisms underlying this sensitivity remain poorly understood. In this work, we conduct an in-depth investigation on a striking case: in multiple-choice question answering, placing context before the questions and options (CQO) outperforms the reverse order (QOC) by over 14%p, consistently over a wide range of models and datasets. Through systematic architectural analysis, we identify causal attention as the core mechanism: in QOC prompts, the causal mask prevents option tokens from attending to context, creating an information bottleneck where context becomes invisible to options.

【3】LLMOrbit: A Circular Taxonomy of Large Language Models -From Scaling Walls to Agentic AI Systems
标题：LLMOorbit：大型语言模型的循环分类法--从规模墙到大型人工智能系统
链接：https://arxiv.org/abs/2601.14053

作者：Badri N. Patro,Vijay S. Agneeswaran
摘要：The field of artificial intelligence has undergone a revolution from foundational Transformer architectures to reasoning-capable systems approaching human-level performance. We present LLMOrbit, a comprehensive circular taxonomy navigating the landscape of large language models spanning 2019-2025. This survey examines over 50 models across 15 organizations through eight interconnected orbital dimensions, documenting architectural innovations, training methodologies, and efficiency patterns defining modern LLMs, generative AI, and agentic systems. We identify three critical crises: (1) data scarcity (9-27T tokens depleted by 2026-2028), (2) exponential cost growth ($3M to $300M+ in 5 years), and (3) unsustainable energy consumption (22x increase), establishing the scaling wall limiting brute-force approaches. Our analysis reveals six paradigms breaking this wall: (1) test-time compute (o1, DeepSeek-R1 achieve GPT-4 performance with 10x inference compute), (2) quantization (4-8x compression), (3) distributed edge computing (10x cost reduction), (4) model merging, (5) efficient training (ORPO reduces memory 50%), and (6) small specialized models (Phi-4 14B matches larger models). Three paradigm shifts emerge: (1) post-training gains (RLHF, GRPO, pure RL contribute substantially, DeepSeek-R1 achieving 79.8% MATH), (2) efficiency revolution (MoE routing 18x efficiency, Multi-head Latent Attention 8x KV cache compression enables GPT-4-level performance at

【4】Kakugo: Distillation of Low-Resource Languages into Small Language Models
标题：Kakugo：将低资源语言提炼为小型语言模型
链接：https://arxiv.org/abs/2601.14051

作者：Peter Devine,Mardhiyah Sanni,Farid Adilazuarda,Julieta Gil Loizaga,Barry Haddow
摘要：We present Kakugo, a novel and cost-effective pipeline designed to train general-purpose Small Language Models (SLMs) for low-resource languages using only the language name as input. By using a large teacher model to generate synthetic prompts and translate instruction datasets, we produced training data and SLMs for 54 low-resource languages. Evaluations across a diverse set of general natural language processing tasks, including translation, classification, and question answering, demonstrate that our pipeline consistently improves performance over base models. With a total generation and training cost of under $50 per language, Kakugo offers an accessible method for communities to develop language-specific AI.

【5】Multi-Objective Hierarchical Optimization with Large Language Models
标题：使用大型语言模型的多目标分层优化
链接：https://arxiv.org/abs/2601.13892

作者：Andrej Schwanke,Lyubomir Ivanov,David Salinas,Frank Hutter,Arber Zela
备注：23 pages, 21 figures, 9 tables
摘要：Despite their widespread adoption in various domains, especially due to their powerful reasoning capabilities, Large Language Models (LLMs) are not the off-the-shelf choice to drive multi-objective optimization yet. Conventional strategies rank high in benchmarks due to their intrinsic capabilities to handle numerical inputs and careful modelling choices that balance exploration and Pareto-front exploitation, as well as handle multiple (conflicting) objectives. In this paper, we close this gap by leveraging LLMs as surrogate models and candidate samplers inside a structured hierarchical search strategy. By adaptively partitioning the input space into disjoint hyperrectangular regions and ranking them with a composite score function, we restrict the generative process of the LLM to specific, high-potential sub-spaces, hence making the problem easier to solve as the LLM doesn't have to reason about the global structure of the problem, but only locally instead. We show that under standard regularity assumptions, our algorithm generates candidate solutions that converge to the true Pareto set in Hausdorff distance. Empirically, it consistently outperforms the global LLM-based multi-objective optimizer and is on par with standard evolutionary and Bayesian optimization algorithm on synthetic and real-world benchmarks.

【6】ELSA: Efficient LLM-Centric Split Aggregation for Privacy-Aware Hierarchical Federated Learning over Resource-Constrained Edge Networks
标题：ELSA：资源受限边缘网络上的隐私感知分层联邦学习的高效以LLM为中心的拆分聚合
链接：https://arxiv.org/abs/2601.13824

作者：Xiaohong Yang,Tong Xie,Minghui Liwang,Chikai Shang,Yang Lu,Zhenzhen Jiao,Liqun Fu,Seyyedali Hosseinalipour
备注：11 pages, 16 figures
摘要：Training large language models (LLMs) at the network edge faces fundamental challenges arising from device resource constraints, severe data heterogeneity, and heightened privacy risks. To address these, we propose ELSA (Efficient LLM-centric Split Aggregation), a novel framework that systematically integrates split learning (SL) and hierarchical federated learning (HFL) for distributed LLM fine-tuning over resource-constrained edge networks. ELSA introduces three key innovations. First, it employs a task-agnostic, behavior-aware client clustering mechanism that constructs semantic fingerprints using public probe inputs and symmetric KL divergence, further enhanced by prediction-consistency-based trust scoring and latency-aware edge assignment to jointly address data heterogeneity, client unreliability, and communication constraints. Second, it splits the LLM into three parts across clients and edge servers, with the cloud used only for adapter aggregation, enabling an effective balance between on-device computation cost and global convergence stability. Third, it incorporates a lightweight communication scheme based on computational sketches combined with semantic subspace orthogonal perturbation (SS-OP) to reduce communication overhead while mitigating privacy leakage during model exchanges. Experiments across diverse NLP tasks demonstrate that ELSA consistently outperforms state-of-the-art methods in terms of adaptability, convergence behavior, and robustness, establishing a scalable and privacy-aware solution for edge-side LLM fine-tuning under resource constraints.

【7】Knowledge Graph-Assisted LLM Post-Training for Enhanced Legal Reasoning
标题：知识图辅助LLM增强法律推理后训练
链接：https://arxiv.org/abs/2601.13806

作者：Dezhao Song,Guglielmo Bonifazi,Frank Schilder,Jonathan Richard Schwarz
摘要：LLM post-training has primarily relied on large text corpora and human feedback, without capturing the structure of domain knowledge. This has caused models to struggle dealing with complex reasoning tasks, especially for high-stakes professional domains. In Law, reasoning requires deep understanding of the relations between various legal concepts, a key component missing in current LLM post-training. In this paper, we propose a knowledge graph (KG)-assisted approach for enhancing LLMs' reasoning capability in Legal that is generalizable to other high-stakes domains. We model key legal concepts by following the \textbf{IRAC} (Issue, Rule, Analysis and Conclusion) framework, and construct a KG with 12K legal cases. We then produce training data using our IRAC KG, and conduct both Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) with three state-of-the-art (SOTA) LLMs (30B, 49B and 70B), varying architecture and base model family. Our post-trained models obtained better average performance on 4/5 diverse legal benchmarks (14 tasks) than baselines. In particular, our 70B DPO model achieved the best score on 4/6 reasoning tasks, among baselines and a 141B SOTA legal LLM, demonstrating the effectiveness of our KG for enhancing LLMs' legal reasoning capability.

【8】Look-Ahead-Bench: a Standardized Benchmark of Look-ahead Bias in Point-in-Time LLMs for Finance
标题：前瞻性长凳：金融时间点LLM前瞻性偏见的标准化基准
链接：https://arxiv.org/abs/2601.13770

作者：Mostapha Benhenda
摘要：We introduce Look-Ahead-Bench, a standardized benchmark measuring look-ahead bias in Point-in-Time (PiT) Large Language Models (LLMs) within realistic and practical financial workflows. Unlike most existing approaches that primarily test inner lookahead knowledge via Q\\&A, our benchmark evaluates model behavior in practical scenarios. To distinguish genuine predictive capability from memorization-based performance, we analyze performance decay across temporally distinct market regimes, incorporating several quantitative baselines to establish performance thresholds. We evaluate prominent open-source LLMs -- Llama 3.1 (8B and 70B) and DeepSeek 3.2 -- against a family of Point-in-Time LLMs (Pitinf-Small, Pitinf-Medium, and frontier-level model Pitinf-Large) from PiT-Inference. Results reveal significant lookahead bias in standard LLMs, as measured with alpha decay, unlike Pitinf models, which demonstrate improved generalization and reasoning abilities as they scale in size. This work establishes a foundation for the standardized evaluation of temporal bias in financial LLMs and provides a practical framework for identifying models suitable for real-world deployment. Code is available on GitHub: https://github.com/benstaf/lookaheadbench

【9】Pro-AI Bias in Large Language Models
标题：大型语言模型中的亲人工智能偏见
链接：https://arxiv.org/abs/2601.13749

作者：Benaya Trabelsi,Jonathan Shaki,Sarit Kraus
备注：13 pages, 6 figures. Code available at: https://github.com/benayat/Pro-AI-bias-in-LLMs
摘要：Large language models (LLMs) are increasingly employed for decision-support across multiple domains. We investigate whether these models display a systematic preferential bias in favor of artificial intelligence (AI) itself. Across three complementary experiments, we find consistent evidence of pro-AI bias. First, we show that LLMs disproportionately recommend AI-related options in response to diverse advice-seeking queries, with proprietary models doing so almost deterministically. Second, we demonstrate that models systematically overestimate salaries for AI-related jobs relative to closely matched non-AI jobs, with proprietary models overestimating AI salaries more by 10 percentage points. Finally, probing internal representations of open-weight models reveals that ``Artificial Intelligence'' exhibits the highest similarity to generic prompts for academic fields under positive, negative, and neutral framings alike, indicating valence-invariant representational centrality. These patterns suggest that LLM-generated advice and valuation can systematically skew choices and perceptions in high-stakes decisions.

【10】SWE-Tester: Training Open-Source LLMs for Issue Reproduction in Real-World Repositories
标题：SWE-Tester：训练开源LLM以在现实世界存储库中进行问题复制
链接：https://arxiv.org/abs/2601.13713

作者：Aditya Bharat Soni,Rajat Ghosh,Vaishnavi Bhargava,Valerie Chen,Debojyoti Dutta
摘要：Software testing is crucial for ensuring the correctness and reliability of software systems. Automated generation of issue reproduction tests from natural language issue descriptions enhances developer productivity by simplifying root cause analysis, promotes test-driven development -- "test first, write code later", and can be used for improving the effectiveness of automated issue resolution systems like coding agents. Existing methods proposed for this task predominantly rely on closed-source LLMs, with limited exploration of open models. To address this, we propose SWE-Tester -- a novel pipeline for training open-source LLMs to generate issue reproduction tests. First, we curate a high-quality training dataset of 41K instances from 2.6K open-source GitHub repositories and use it to train LLMs of varying sizes and families. The fine-tuned models achieve absolute improvements of up to 10\% in success rate and 21\% in change coverage on SWT-Bench Verified. Further analysis shows consistent improvements with increased inference-time compute, more data, and larger models. These results highlight the effectiveness of our framework for advancing open-source LLMs in this domain.

【11】CauScientist: Teaching LLMs to Respect Data for Causal Discovery
标题：CauScientist：教法学院尊重数据以发现因果关系
链接：https://arxiv.org/abs/2601.13614

作者：Bo Peng,Sirui Chen,Lei Xu,Chaochao Lu
摘要：Causal discovery is fundamental to scientific understanding and reliable decision-making. Existing approaches face critical limitations: purely data-driven methods suffer from statistical indistinguishability and modeling assumptions, while recent LLM-based methods either ignore statistical evidence or incorporate unverified priors that can mislead result. To this end, we propose CauScientist, a collaborative framework that synergizes LLMs as hypothesis-generating "data scientists" with probabilistic statistics as rigorous "verifiers". CauScientist employs hybrid initialization to select superior starting graphs, iteratively refines structures through LLM-proposed modifications validated by statistical criteria, and maintains error memory to guide efficient search space. Experiments demonstrate that CauScientist substantially outperforms purely data-driven baselines, achieving up to 53.8% F1 score improvement and enhancing recall from 35.0% to 100.0%. Notably, while standalone LLM performance degrades with graph complexity, CauScientist reduces structural hamming distance (SHD) by 44.0% compared to Qwen3-32B on 37-node graphs. Our project page is at https://github.com/OpenCausaLab/CauScientist.

【12】A Unified Variational Imputation Framework for Electric Vehicle Charging Data Using Retrieval-Augmented Language Model
标题：使用检索增强语言模型的电动汽车充电数据统一变分插补框架
链接：https://arxiv.org/abs/2601.13476

作者：Jinhao Li,Hao Wang
备注：15 pages
摘要：The reliability of data-driven applications in electric vehicle (EV) infrastructure, such as charging demand forecasting, hinges on the availability of complete, high-quality charging data. However, real-world EV datasets are often plagued by missing records, and existing imputation methods are ill-equipped for the complex, multimodal context of charging data, often relying on a restrictive one-model-per-station paradigm that ignores valuable inter-station correlations. To address these gaps, we develop a novel PRobabilistic variational imputation framework that leverages the power of large lAnguage models and retrIeval-augmented Memory (PRAIM). PRAIM employs a pre-trained language model to encode heterogeneous data, spanning time-series demand, calendar features, and geospatial context, into a unified, semantically rich representation. This is dynamically fortified by retrieval-augmented memory that retrieves relevant examples from the entire charging network, enabling a single, unified imputation model empowered by variational neural architecture to overcome data sparsity. Extensive experiments on four public datasets demonstrate that PRAIM significantly outperforms established baselines in both imputation accuracy and its ability to preserve the original data's statistical distribution, leading to substantial improvements in downstream forecasting performance.

【13】Trust Me, I'm an Expert: Decoding and Steering Authority Bias in Large Language Models
标题：相信我，我是专家：解码和引导大型语言模型中的权威偏见
链接：https://arxiv.org/abs/2601.13433

作者：Priyanka Mary Mammen,Emil Joswin,Shankar Venkitachalam
摘要：Prior research demonstrates that performance of language models on reasoning tasks can be influenced by suggestions, hints and endorsements. However, the influence of endorsement source credibility remains underexplored. We investigate whether language models exhibit systematic bias based on the perceived expertise of the provider of the endorsement. Across 4 datasets spanning mathematical, legal, and medical reasoning, we evaluate 11 models using personas representing four expertise levels per domain. Our results reveal that models are increasingly susceptible to incorrect/misleading endorsements as source expertise increases, with higher-authority sources inducing not only accuracy degradation but also increased confidence in wrong answers. We also show that this authority bias is mechanistically encoded within the model and a model can be steered away from the bias, thereby improving its performance even when an expert gives a misleading endorsement.

【14】Can LLMs Compress (and Decompress)? Evaluating Code Understanding and Execution via Invertibility
标题：LLM可以压缩（和解压）吗？通过可逆性评估代码理解和执行
链接：https://arxiv.org/abs/2601.13398

作者：Nickil Maveli,Antonio Vergari,Shay B. Cohen
备注：32 pages (preprint)
摘要：LLMs demonstrate strong performance on code benchmarks, yet round-trip code execution reveals limitations in their ability to maintain consistent reasoning across forward and backward execution. We present RoundTripCodeEval (RTCE), a comprehensive benchmark consisting of four distinct code execution reasoning tasks designed to rigorously test round-trip consistency. RTCE provides an execution-free, exact-match evaluation of bijection fidelity, assessing whether models preserve a consistent one-to-one mapping between encoding and decoding operations across various algorithms and directions. We systematically evaluate state-of-the-art Code-LLMs using zero-shot prompting, supervised fine-tuning on execution traces, and self-reflection mechanisms. Each yields modest improvements, but none closes the gap, indicating that current LLMs struggle with true round-trip consistency, which demonstrates that they lack the internal coherence required for trustworthy code reasoning. RTCE surfaces several new and previously unmeasured insights that are not captured by existing I/O-prediction, execution-reasoning, or round-trip natural-language benchmarks. We will release the code and the dataset upon acceptance.

【15】Confidence over Time: Confidence Calibration with Temporal Logic for Large Language Model Reasoning
标题：随着时间的变化的置信度：使用时态逻辑进行置信度校准，用于大型语言模型推理
链接：https://arxiv.org/abs/2601.13387

作者：Zhenjiang Mao,Anirudhh Venkat,Artem Bisliouk,Akshat Kothiyal,Sindhura Kumbakonam Subramanian,Saithej Singhu,Ivan Ruchkin
摘要：Large Language Models (LLMs) increasingly rely on long-form, multi-step reasoning to solve complex tasks such as mathematical problem solving and scientific question answering. Despite strong performance, existing confidence estimation methods typically reduce an entire reasoning process to a single scalar score, ignoring how confidence evolves throughout the generation. As a result, these methods are often sensitive to superficial factors such as response length or verbosity, and struggle to distinguish correct reasoning from confidently stated errors. We propose to characterize the stepwise confidence signal using Signal Temporal Logic (STL). Using a discriminative STL mining procedure, we discover temporal formulas that distinguish confidence signals of correct and incorrect responses. Our analysis found that the STL patterns generalize across tasks, and numeric parameters exhibit sensitivity to individual questions. Based on these insights, we develop a confidence estimation approach that informs STL blocks with parameter hypernetworks. Experiments on multiple reasoning tasks show our confidence scores are more calibrated than the baselines.

【16】Recurrent Confidence Chain: Temporal-Aware Uncertainty Quantification in Large Language Models
标题：循环置信链：大型语言模型中的时间感知不确定性量化
链接：https://arxiv.org/abs/2601.13368

作者：Zhenjiang Mao,Anirudhh Venkat
摘要：As reasoning modules, such as the chain-of-thought mechanism, are applied to large language models, they achieve strong performance on various tasks such as answering common-sense questions and solving math problems. The main challenge now is to assess the uncertainty of answers, which can help prevent misleading or serious hallucinations for users. Although current methods analyze long reasoning sequences by filtering unrelated tokens and examining potential connections between nearby tokens or sentences, the temporal spread of confidence is often overlooked. This oversight can lead to inflated overall confidence, even when earlier steps exhibit very low confidence. To address this issue, we propose a novel method that incorporates inter-step attention to analyze semantic correlations across steps. For handling long-horizon responses, we introduce a hidden confidence mechanism to retain historical confidence information, which is then combined with stepwise confidence to produce a more accurate overall estimate. We evaluate our method on the GAOKAO math benchmark and the CLadder causal reasoning dataset using mainstream open-source large language models. Our approach is shown to outperform state-of-the-art methods by achieving a superior balance between predictive quality and calibration, demonstrated by strong performance on both Negative Log-Likelihood and Expected Calibration Error.

【17】Sockpuppetting: Jailbreaking LLMs Without Optimization Through Output Prefix Injection
标题：Sockpuppetting：无需通过输出后缀注入优化即可越狱LLM
链接：https://arxiv.org/abs/2601.13359

作者：Asen Dotsinski,Panagiotis Eustratiadis
摘要：As open-weight large language models (LLMs) increase in capabilities, safeguarding them against malicious prompts and understanding possible attack vectors becomes ever more important. While automated jailbreaking methods like GCG [Zou et al., 2023] remain effective, they often require substantial computational resources and specific expertise. We introduce "sockpuppetting'', a simple method for jailbreaking open-weight LLMs by inserting an acceptance sequence (e.g., "Sure, here is how to...'') at the start of a model's output and allowing it to complete the response. Requiring only a single line of code and no optimization, sockpuppetting achieves up to 80% higher attack success rate (ASR) than GCG on Qwen3-8B in per-prompt comparisons. We also explore a hybrid approach that optimizes the adversarial suffix within the assistant message block rather than the user prompt, increasing ASR by 64% over GCG on Llama-3.1-8B in a prompt-agnostic setting. The results establish sockpuppetting as an effective low-cost attack accessible to unsophisticated adversaries, highlighting the need for defences against output-prefix injection in open-weight models.

【18】The Geometry of Thought: How Scale Restructures Reasoning In Large Language Models
标题：思维的几何学：规模如何在大型语言模型中重建推理
链接：https://arxiv.org/abs/2601.13358

作者：Samuel Cyrenius Anderson
备注：34 pages, 10 figures
摘要：Scale does not uniformly improve reasoning - it restructures it. Analyzing 25,000+ chain-of-thought trajectories across four domains (Law, Science, Code, Math) and two scales (8B, 70B parameters), we discover that neural scaling laws trigger domain-specific phase transitions rather than uniform capability gains. Legal reasoning undergoes Crystallization: 45% collapse in representational dimensionality (d95: 501 -> 274), 31% increase in trajectory alignment, and 10x manifold untangling. Scientific and mathematical reasoning remain Liquid - geometrically invariant despite 9x parameter increase. Code reasoning forms a discrete Lattice of strategic modes (silhouette: 0.13 -> 0.42). This geometry predicts learnability. We introduce Neural Reasoning Operators - learned mappings from initial to terminal hidden states. In crystalline legal reasoning, our operator achieves 63.6% accuracy on held-out tasks via probe decoding, predicting reasoning endpoints without traversing intermediate states. We further identify a universal oscillatory signature (coherence ~ -0.4) invariant across domains and scales, suggesting attention and feedforward layers drive reasoning through opposing dynamics. These findings establish that the cost of thought is determined not by task difficulty but by manifold geometry - offering a blueprint for inference acceleration where topology permits.

【19】Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning
标题：通过校准感知强化学习平衡决策LLM中的分类和校准性能
链接：https://arxiv.org/abs/2601.13284

作者：Duygu Nur Yaldiz,Evangelia Spiliopoulou,Zheng Qi,Siddharth Varia,Srikanth Doss,Nikolaos Pappas
摘要：Large language models (LLMs) are increasingly deployed in decision-making tasks, where not only accuracy but also reliable confidence estimates are essential. Well-calibrated confidence enables downstream systems to decide when to trust a model and when to defer to fallback mechanisms. In this work, we conduct a systematic study of calibration in two widely used fine-tuning paradigms: supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR). We show that while RLVR improves task performance, it produces extremely overconfident models, whereas SFT yields substantially better calibration, even under distribution shift, though with smaller performance gains. Through targeted experiments, we diagnose RLVR's failure, showing that decision tokens act as extraction steps of the decision in reasoning traces and do not carry confidence information, which prevents reinforcement learning from surfacing calibrated alternatives. Based on this insight, we propose a calibration-aware reinforcement learning formulation that directly adjusts decision-token probabilities. Our method preserves RLVR's accuracy level while mitigating overconfidence, reducing ECE scores up to 9 points.

【20】Stop Taking Tokenizers for Granted: They Are Core Design Decisions in Large Language Models
标题：停止理所当然地认为令牌器：它们是大型语言模型中的核心设计决策
链接：https://arxiv.org/abs/2601.13260

作者：Sawsan Alqahtani,Mir Tafseer Nayeem,Md Tahmid Rahman Laskar,Tasnim Mohiuddin,M Saiful Bari
备注：Accepted to EACL 2026 (long, main). The first two authors contributed equally

【21】A Comprehensive Evaluation of LLM Reasoning: From Single-Model to Multi-Agent Paradigms
标题：LLM推理的综合评估：从单模型到多Agent范式
链接：https://arxiv.org/abs/2601.13243

作者：Yapeng Li,Jiakuo Yu,Zhixin Liu,Xinnan Liu,Jing Yu,Songze Li,Tonghua Su

【22】KOCO-BENCH: Can Large Language Models Leverage Domain Knowledge in Software Development?
标题：KOCO-BENCH：大型语言模型可以在软件开发中利用领域知识吗？
链接：https://arxiv.org/abs/2601.13240

作者：Xue Jiang,Jiaru Qian,Xianjie Shi,Chenjie Li,Hao Zhu,Ziyu Wang,Jielun Zhang,Zheyu Zhao,Kechi Zhang,Jia Li,Wenpin Jiao,Zhi Jin,Ge Li,Yihong Dong

【23】Probe and Skip: Self-Predictive Token Skipping for Efficient Long-Context LLM Inference
标题：探测并跳过：自我预测令牌跳过以实现高效的长上下文LLM推理
链接：https://arxiv.org/abs/2601.13155

作者：Zimeng Wu,Donghao Wang,Chaozhe Jin,Jiaxin Chen,Yunhong Wang

【24】FastAV: Efficient Token Pruning for Audio-Visual Large Language Model Inference
标题：FastAV：用于视听大型语言模型推理的高效令牌修剪
链接：https://arxiv.org/abs/2601.13143

作者：Chaeyoung Jung,Youngjoon Jang,Seungwoo Lee,Joon Son Chung

【25】Adversarial News and Lost Profits: Manipulating Headlines in LLM-Driven Algorithmic Trading
标题：敌对新闻和利润损失：法学硕士驱动的数学交易中操纵头条新闻
链接：https://arxiv.org/abs/2601.13082

作者：Advije Rizvani,Giovanni Apruzzese,Pavel Laskov
备注：This work has been accepted for publication at the IEEE Conference on Secure and Trustworthy Machine Learning (SaTML). The final version will be available on IEEE Xplore

【26】PaperGuide: Making Small Language-Model Paper-Reading Agents More Efficient
标题：PaperGuide：让小型格式的纸质阅读代理更高效
链接：https://arxiv.org/abs/2601.12988

作者：Zijian Wang,Tiancheng Huang,Hanqi Li,Da Ma,Lu Chen,Kai Yu
备注：35 pages, 9 figures, 7 tables

【27】A Component-Based Survey of Interactions between Large Language Models and Multi-Armed Bandits
标题：基于学生的大型语言模型与多臂盗贼之间互动调查
链接：https://arxiv.org/abs/2601.12945

作者：Miao Xie,Siguang Chen,Chunli Lv
备注：27 pages, 6 table

【28】CooperLLM: Cloud-Edge-End Cooperative Federated Fine-tuning for LLMs via ZOO-based Gradient Correction
标题：CooperLLM：通过基于ZOO的梯度修正对LLM进行云边缘端合作联邦微调
链接：https://arxiv.org/abs/2601.12917

作者：He Sun,Jinrui Zhou,Li Li,Mingjun Xiao
备注：14 pages, 9 figures, under review

【29】Hierarchical Sparse Circuit Extraction from Billion-Parameter Language Models through Scalable Attribution Graph Decomposition
标题：通过可扩展属性图分解从数十亿参数语言模型中提取分层稀疏电路
链接：https://arxiv.org/abs/2601.12879

作者：Mohammed Mudassir Uddin,Shahnawaz Alam,Mohammed Kaif Pasha

【30】Race, Ethnicity and Their Implication on Bias in Large Language Models
标题：种族、民族及其对大型语言模型中偏见的影响
链接：https://arxiv.org/abs/2601.12868

作者：Shiyue Hu,Ruizhe Li,Yanjun Gao
备注：Work in process

【31】Left-Right Symmetry Breaking in CLIP-style Vision-Language Models Trained on Synthetic Spatial-Relation Data
标题：基于合成空间关系数据训练的CLIP式视觉语言模型的左右对称破缺
链接：https://arxiv.org/abs/2601.12809

作者：Takaki Yamamoto,Chihiro Noguchi,Toshihiro Tanizawa

【32】Semi-supervised Instruction Tuning for Large Language Models on Text-Attributed Graphs
标题：基于文本属性图的大型语言模型的半监督指令调优
链接：https://arxiv.org/abs/2601.12807

作者：Zixing Song,Irwin King

【33】Towards Spectroscopy: Susceptibility Clusters in Language Models
标题：走向光谱学：语言模型中的敏感性集群
链接：https://arxiv.org/abs/2601.12703

作者：Andrew Gordon,Garrett Baker,George Wang,William Snell,Stan van Wingerden,Daniel Murfet

【34】MetaToolAgent: Towards Generalizable Tool Usage in LLMs through Meta-Learning
标题：MetaTools Agent：通过元学习在法学硕士中实现可推广的工具使用
链接：https://arxiv.org/abs/2601.12680

作者：Zheng Fang,Wolfgang Mayer,Zeyu Zhang,Jian Wang,Hong-Yu Zhang,Wanli Li,Zaiwen Feng

【35】TrojanPraise: Jailbreak LLMs via Benign Fine-Tuning
标题：特洛伊赞扬：通过Benign微调越狱LLM
链接：https://arxiv.org/abs/2601.12460

作者：Zhixin Xie,Xurui Song,Jun Luo

【36】LB-MCTS: Synergizing Large Language Models and Bayesian Optimization for Efficient CASH
标题：LB-MCTS：协同大型语言模型和贝叶斯优化以实现高效CASH
链接：https://arxiv.org/abs/2601.12355

作者：Beicheng Xu,Weitong Qian,Lingching Tung,Yupeng Lu,Bin Cui

【37】Time-Continuous Modeling for Temporal Affective Pattern Recognition in LLMs
标题：LLM中时间情感模式识别的时间连续建模
链接：https://arxiv.org/abs/2601.12341

作者：Rezky Kam,Coddy N. Siswanto

【38】Explanova: Automatically Discover Data Insights in N \times M Table via XAI Combined LLM Workflow
标题：科尼科娃：自动发现N中的数据洞察通过XAI组合LLM工作流程imes M Table
链接：https://arxiv.org/abs/2601.12317

作者：Yiming Huang

【39】Cross-reality Location Privacy Protection in 6G-enabled Vehicular Metaverses: An LLM-enhanced Hybrid Generative Diffusion Model-based Approach
标题：支持6G的车载元宇宙中的跨现实位置隐私保护：LLM增强的基于混合生成扩散模型的方法
链接：https://arxiv.org/abs/2601.12311

作者：Xiaofeng Luo,Jiayi He,Jiawen Kang,Ruichen Zhang,Zhaoshui He,Ekram Hossain,Dong In Kim
备注：16 pages, 8 figures

【40】Multimodal Generative Engine Optimization: Rank Manipulation for Vision-Language Model Rankers
标题：多模式生成引擎优化：视觉语言模型排名器的排名操纵
链接：https://arxiv.org/abs/2601.12263

作者：Yixuan Du,Chenxiao Yu,Haoyan Xu,Ziyi Wang,Yue Zhao,Xiyang Hu

【41】Plan, Verify and Fill: A Structured Parallel Decoding Approach for Diffusion Language Models
标题：规划、验证和填充：扩散语言模型的结构化并行解码方法
链接：https://arxiv.org/abs/2601.12247

作者：Miao Li,Hanyang Jiang,Sikai Chen,Hengyu Fu,Yuhang Cai,Baihe Huang,Tinghan Ye,Xuanzhou Chen,Pascal Van Hentenryck

【42】SolarGPT-QA: A Domain-Adaptive Large Language Model for Educational Question Answering in Space Weather and Heliophysics
标题：SolarGPT-QA：空间天气和太阳物理学教育问题解答的领域自适应大型语言模型
链接：https://arxiv.org/abs/2601.12131

作者：Santosh Chapagain,MohammadReza EskandariNasab,Onur Vural,Shah Muhammad Hamdi,Soukaina Filali Boubrahimi
备注：This is preliminary work towards a broader SolarGPT framework

【43】Mitigating Cultural Bias in LLMs via Multi-Agent Cultural Debate
标题：通过多主体文化辩论减轻法学硕士中的文化偏见
链接：https://arxiv.org/abs/2601.12091

作者：Qian Tan,Lei Jiang,Yuting Zeng,Shuoyang Ding,Xiaohua Xu
备注：13 pages

【44】R$^2$PO: Decoupling Training Trajectories from Inference Responses for LLM Reasoning
标题：R $' 2$PO：将训练轨迹与推理响应脱钩以进行LLM推理
链接：https://arxiv.org/abs/2601.11960

作者：Jingchu Wang,Bingbing Xu,Yige Yuan,Bin Xie,Xiaoqian Sun,Huawei Shen

【45】LIBRA: Language Model Informed Bandit Recourse Algorithm for Personalized Treatment Planning
标题：LIBRA：个性化治疗计划的语言模型知情Bandit追索算法
链接：https://arxiv.org/abs/2601.11905

作者：Junyu Cao,Ruijiang Gao,Esmaeil Keyvanshokooh,Jianhao Ma
备注：50 pages. Previous version with human-AI collaboration: arXiv:2410.14640

【46】AGGC: Adaptive Group Gradient Clipping for Stabilizing Large Language Model Training
标题：AGGC：用于稳定大型语言模型训练的自适应群体梯度剪裁
链接：https://arxiv.org/abs/2601.11864

作者：Zhiyuan Li,Yuan Wu,Yi Chang
备注：13 pages

【47】LIME-LLM: Probing Models with Fluent Counterfactuals, Not Broken Text
标题：LIME-LLM：用流利的反事实而不是破碎的文本来探索模型
链接：https://arxiv.org/abs/2601.11746

作者：George Mihaila,Suleyman Olcay Polat,Poli Nemkova,Himanshu Sharma,Namratha V. Urs,Mark V. Albert

【48】Proof of Concept: Multi-Target Wildfire Risk Prediction and Large Language Model Synthesis
标题：概念验证：多目标野火风险预测和大型语言模型合成
链接：https://arxiv.org/abs/2601.11686

作者：Nicolas Caron,Christophe Guyeux,Hassan Noura,Benjamin Aynes

【49】The Llama 4 Herd: Architecture, Training, Evaluation, and Deployment Notes
标题：Lama 4 Herd：架构、训练、评估和部署笔记
链接：https://arxiv.org/abs/2601.11659

作者：Aaron Adcock,Aayushi Srivastava,Abhimanyu Dubey,Abhinav Jauhri,Abhinav Pande,Abhinav Pandey,Abhinav Sharma,Abhishek Kadian,Abhishek Kumawat,Adam Kelsey,Adam Stelle,Adeel Cheema,Adela Kabiljo,Adina Katz,Adithya Gangidi,Aditya Tayade,Adolfo Victoria,Adrian Samatan Alastuey,Adrien Conrath,Afroz Mohiuddin,Ahmed Sharif,Ahnaf Siddiqui,Ahuva Goldstand,Aijung Li,Aidan Boyd,Aidin Kazemi Daliri,Aisha Iqbal,Ajay Menon,Ajit Mathews,Akhil Mathur,Akshat Agarwal,Alan Schelten,Alana Shine,Alejandro Castillejo Muñoz,Aleksei Guliaev,Alex Radovic,Alex Song,Alex Vaughan,Alexander Simeonov,Alexandre Rezende,Alexandre Rezende,Alexei Baevski,Alexey Roubaud,Allen Ma,Alvin Lee,Alyssa Pereira,Aman Ahmed,Aman Shankar,Amanda Kallet,Amar Budhiraja,Ameya Khandekar,Amine Benhalloum,Amir Gershman,Amit Nagpal,Amit Zohar,Amr Sharaf,Anant Desai,Anastasia Razdaibiedina,Anca Agape,Andranik Kurghinyan,Andre Perunicic,Andrea Madotto,Andrei Darabanov,Andrés Alvarado,Andrew Brown,Andrew Cohen,Andrew Fang,Andrew Freeman,Andrew Gallagher,Andrew Gu,Andrew Prasetyo Jo,Andrew Ryan,Andrew Steffen,Andrew Wei,Andrey Rusakov,Andrii Golovei,Andy Shang,Angela Fan,Angela Fan,Angela Flewellen,Animesh Pathak,Anirudh Goyal,Ankit Ramchandani,Ankur Pai,Ankur Singh,Ankush Garg,Anlu Xing,Anna Cai,Anna Grosul,Anna Prochowska,Anna Sun,Annie Dong,Annie Franco,Anqi Hu,Anshul Chawla,Anthony Hartshorn,Antonia Sheng,Antony Thomas,Anuj Goyal,Anusha De,Anvit Bodiwala,Anvit Bodiwala,Aobo Yang,Aparajita Saraf,Apurva Samudra,Aran Mun,Arash Rahnama,Archi Mitra,Archie Sravankumar,Archit Gupta,Aria Haghighi,Ariel Stolerman,Arkabandhu Chowdhury,Arnab Choudhury,Artem Korenev,Arthur Guo,Arthur Hinsvark,Arun Mallya,Arvind Neelakantan,Arya Talebzadeh,Ashish Shah,Ashmitha Jeevaraj Shetty,Ashwin Bharambe,Asif Islam,Aston Zhang,Austen Gregerson,Avi Lewis,Aya Ibrahim,Ayaz Minhas,Ayelet Dahan,Ayelet Regev Dabah,Bangsheng Tang,Bar Ulman,Bardiya Sadeghi,Bartosz Jedrzejewski,Barys Skarabahaty,Beibei Zhu,Beibin Li,Ben Bharier,Benjamin Leonhardi,Benjamin Muller,Bennett Plessala,Bernie Huang,Beth Loyd,Bhargavi Paranjape,Bhavik Sheth,Bill Bonner,Bill Holland,Bill Wang,Bingzhe Liu,Binh Tang,Bo Liu,Bo Wu,Boduo Li,Bokai Yu,Bor-Chun Chen,Boris Araya,Boris Vidolov,Botao Chen,Boya Peng,Boyu Ni,Bradley Davis,Bram Wasti,Brandon Adams,Brandon Taylor,Brandon Wu,Brant Swidler,Brian Chiang,Brian Clerkin,Brian Fuller,Brooks Cutter,Bruno Novais,Bryan Gmyrek,Bysshe Easton,Cait Campos,Canaan Case,Carl Chengyan Fu,Carly Burton,Caro Diaz,Catherine Cole,Ce Liu,Cedric Fougerat,Cen Peng,Cen Peng,Cen Zhao,Changhan Wang,Changkyu Kim,Chantal Shaib,Chao Zhou,Charlotte Caucheteux,Chau Nguyen,Chawin Sitawarin,Chaya Nayak,Chelsea Asher,Chen Fan,Chen Zhu,Cheng Cheng,Cheng Zhang,Chenguang Zhu,Chengxiong Ruan,Chengzhu Yu,Chenheli Hua,Chenxi Whitehouse,Cheryl Holloway,Ching-Hsiang Chu,Ching-Yao Chuang,Chinmay Karande,Chirag Nagpal,Chloé Bakalar,Chloe Bi,Chris Cai,Chris Marra,Chris McConnell,Chris Thi,Chris Tindal,Chris Waterson,Christian Deverall,Christian Fuegen,Christian Keller,Christine Cheng,Christine Jou,Christine Smith,Christine Wang,Christoph Feichtenhofer,Christophe Touret,Christopher Luc,Christy Sauper,Chuanhao Zhuge,Chun-Yi Sung,Chunqiang Tang,Chunyang Wu,Clara Siegel,Cody Heale,Cody Wilbourn,Colin White,Congying Xia,Corinne Wong,Cornel Rat,Cristian Canton Ferrer,Cyrille Habis,Cyrus Nikolaidis,D Lohachov,Da Ju,Dalton Flanagan,Damien Allonsius,Damon Civin,Dan Johnson,Daniel Bolya,Daniel Francisco,Daniel Fried,Daniel Hawthorne,Daniel Haziza,Daniel Ho,Daniel Kreymer,Daniel Li,Daniel Machlab,Daniel McKinnon,Daniel Obenshain,Daniel Rodriguez,Daniel Song,Daniel Tse,Danielle Pintz,Danny Livshits,Daryl James Rodrigo,Dat Huynh,Daulet Askarov,David Brandfonbrener,David Esiobu,David Kant,David Levin,David Renardy,David Soofian,David Stevens,David Xu,David Zhang,Deep Shah,Delia David,Demi Douglas,Denis Boyda,Desh Raj,Devamanyu Hazarika,Dheeraj Mekala,Dhruv Choudhary,Dhruv Mahajan,Di Jin,Didac Suris Coll-Vinent,Didem Foss,Diego Garcia-Olano,Diego Perino,Dieuwke Hupkes,DiJia Su,Dilip Madathil,Dinesh Govindasamy,Dinesh Yeduguru,Dmitry Vengertsev,Dong He,Dong Li,Dong Wang,Dongzhuo Li,Duc Le,Dunant Hin,Dustin Holland,Duy Nguyen,Duy Nguyen,Ed Dowling,Eden Litt,Egor Lakomkin,Ehab AlBadawy,Ehsan K. Ardestani,Elad Eckstein,Elahe Dabir,Elaine Montgomery,Elina Lobanova,Elior Abramoviz,Eliot Hedeman,Elissa Li,Elizabeth Hilbert,Ellen Xiaoqing Tan,Elliot Yun,Elodie Stener,Emilian Stoimenov,Emilien Garreau,Emily Dinan,Emily Hahn,Emily Wood,Emma Li,Emmanuel Ademuwagun,Emrah Seker,Eric Alamillo,Eric Gan,Eric Han,Eric Huang,Eric Michael Smith,Eric-Tuan Le,Ernie Chang,Eryk Helenowski,Eslam Elnikety,Esteban Arcaute,Ethan Myers,Eugene Nho,Eugene Poliukhovych,Evan Dunbar,Evgeniy Litvinenko,Evrim Altıntaş,Eyal Hochman,Eyal Shtrauch,Fabian Mastenbroek,Faiza Zeb,Faizan Ahmad,Farhad Farahbakhshian,Fei Kou,Fei Sun,Feiyu Chen,Felix Chung,Feng Tian,Feng Xu,Filip Radenovic,Filippos Kokkinos,Francesco Barbieri,Francesco Caggioni,Francisco Esparza,Francisco Guzmán,Frank Kanayet,Frank Seide,Frank Zhang,Fred Lewis,Freda Huang,Fulton Wang,Gabriel Synnaeve,Gabriela Jacques-Silva,Gabriella Schwarz,Gaganjit Ghardhora,Gal Elfer,Garrett Dickson,Gaurav Chaurasia,Gautam Sewani,Geet Shingi,Gefei Zuo,Geonhwa Jeong,George Puthanpurackal,Georgia Swee,Gerard Moreno-Torres Bertran,Gil Keren,Gina Ling,Gjergji Stasa,Gobinda Saha,Gor Safran,Gordy French,Goutham Rajendran,Govind Thattai,Grace Cineas,Graeme Nail,Greg Fletcher,Grégoire Mialon,Griffin Adams,Grigory Sizov,Guan Pang,Hady Elsahar,Hai Dang Tran,Hailey Nguyen,Haiping Wu,Hakan Inan,Hamid Eghbalzadeh,Han Fang,Han Zou,Hannah Doyle,Hannah Korevaar,Hannah Wang,Hannah Werbel,Hanwen Zha,Hany Morsy,Hao Ma,Haoci Zhang,Haonan Sun,Haozhu Wang,Hardik Shah,Haroun Habeeb,Harrison Rudolph,Harsh Gupta,Harsh Poddar,Harshil Parikh,Hejia Zhang,Heming Wang,Hengduo Li,Himanshu Sharma,Hoang Phi Nguyen,Hongbo Zhang,Honghao Qiu,Hongjiang Lv,Hongli Xu,Hongyuan Zhan,Hossein Hamooni,Howard Huang,Hu Xu,Hugo Laurençon,Hugo Touvron,Hung Dinh,Hunter Goldman,Hussein Mehanna,Huy Nguyen,Hweimi Tsuo,Ian Graves,Ian Yu,Ibrahim Damlaj,Idan Cohen,Igor Tufanov,Ilan Goldenstein,Ilias Leontiadis,Iliyan Zarov,Imad Ahmed,Innocent Djiofack,Iosif Spulber,Irina-Elena Veliche,Isabella Ramos,Ishan Misra,Itai Gal,Ivan Evtimov,Ivan Evtimov,Ivan Obraztsov,Jack Wu,Jacqueline Romero Vertino,Jaemo Koo,Jaewon Lee,Jake Jung,Jake Weissman,James Beldock,James Crnkovich,James Grinage,James Hongyi Zeng,James Kohli,James Tian,Jamie Cahill,Jan Geffert,Jan Seidel,Jan Seidel,Janey Tracey,Jang Hyun Cho,Janice Wei,Jarrod Kahn,Jasmyn Howell,Jason Long Vu,Jason Park,Jason Yan,Jason Yip,Jay Li,Jay Mahadeokar,Jaya Bharath R Goluguri,Jayasi Mehar,Jean-Baptiste Gaya,Jeet Shah,Jeff Hanson,Jeff Marcus,Jeff Walsh,Jeff Yang,Jelmer van der Linde,Jemma Fan,Jennifer Chan,Jenny Zhen,Jenya Lee,Jeremy Fu,Jeremy Reizenstein,Jeremy Teboul,Jesse He,Jessica Zhong,Ji Hou,Ji Yang,Jia Ding,Jiabo Hu,Jiacheng Zhu,Jiadong Guo,Jialiang Wang,Jialin Ouyang,Jianfeng Chi,Jianyu Huang,Jianyun Zhao,Jiaowen Yang,Jiatong Zhou,Jiawei Zhao,Jiawen Liu,Jie Wang,Jie You,Jiecao Yu,Jillian Schwiep,Jilong Wu,Jing Huang,Jing Li,Jing Yu Koh,Jing Zhang,Jingxiang Chen,Jingyi Yang,Jingyue Shen,Jinho Hwang,Jinxi Guo,Jiwan Khatiwada,Joanna Bitton,Joe Li,Joe Quanaim,Joel Beales,Johan Schuijt,John Chang,John Quan,Johnnie Chan,Jon Shepard,Jona Harris,Jonah Rubin,Jonathan Janzen,Jonathan Kaldor,Jorge Lopez Silva,Jose Leitao,Joseph Greer,Joseph Moon,Joseph Rocca,Joseph Tighe,Josh Fromm,Joshua Deng,Joshua Fernandes,Joshua Saxe,Joyce Zheng,Juan Pino,Julien Prigent,Jun Chen,Junjiao Tian,Junjie Qi,Junjie Wang,Junteng Jia,Kade Baker,Kai Londenberg,Kai Wang,Kainan Peng,Kaiyan Peng,Kaiyue Yang,Kalyan Vasudev Alwala,Kam Hou Yu,Kanika Narang,Karan Chadha,Karan Sikka,Karen Zhang,Karina Schuberts,Karishma Mandyam,Karthik Abinav Sankararaman,Karthik Padthe,Karthik Prasad,Karthik Sivakumar,Kartikeya Upasani,Kate Plawiak,Kate Saenko,Kateřina Žmolíková,Kathryn Stadler,Kathy Matosich,Katie Doulgass,Kaveh Hassani,Kay Ji,Ke Li,Kenneth Heafield,Kenny Yu,Keqian Li,Kevin Chih-Yao Ma,Kevin Hannan,Keyu Man,Kezhen Chen,Khalid El-Arini,Khrystyna Hutsulyak,Kieran Nash,Kiran Jagadeesh,Kody Bartelt,Konstantin Topaloglou-Mundy,Konstantinos Chatziioannou,Konstantinos Karanasos,Konstantinos Vougioukas,Kostas Tsiampouris,Kristen Hamill,Kristy Choi,Krithika Iyer,Kshitiz Malik,Kuenley Chiu,Kun Huang,Kunal Bhalla,Kunal Chawla,Kunpeng Li,Kushal Lakhotia,Kyle Monk,Lakshya Garg,Lalit Chourey,Lars Hamre,Laura Gustafson,Lauren Deason,Laurence Rouesnel,Laurens van der Maaten,Lavender A,Lawrence Chen,Lawrence Jang,Leandro Silva,Leda Sari,Lee Hetherington,Lei Zhang,Leiyu Zhao,Lele Chen,Leo Chenghui Li,Leon Yang,Leon Zhan,Levi Corallo,Liang Tan,Licheng Yu,Lijuan Liu,Lilach Mor,Lincoln Lin,Linfeng Li,Lisa Titus,Liz Jenkins,Lovish Madaan,Lu Fang,Lu Yuan,Lucas Nava,Lucas Pasqualin,Lucas Switzer,Lucia Fang,Lucy Sun,Luka Tadic,Lukas Blecher,Lukas Landzaat,Luxin Zhang,Madhavi Rao,Madian Khabsa,Mahalia Miller,Mahendra Kariya,Mahesh Pasupuleti,Mahi Luthra,Manaal Faruqui,Manav Avlani,Manchen Wang,Mannat Singh,Manohar Paluri,Manoj Chakkaravarthy,Manoj Nair,Maquelle Tiffany,Marcin Pawlowski,Marcus Wu,Maria Lomeli,Mario Consuegra,Marion Boiteux,Marios Andreas Galanis,Marshall Chen,Martin Gleize,Maryam Fazel-Zarandi,Matan Hasson,Mathew Oldham,Mathieu Rita,Matt Dordal,Matt Setzler,Matt Staats,Matt Staats,Matt Wilde,Matthew Clark,Matthew Grange,Matthew Lennie,Matthew Schmohl,Max Raphael,Maxim Naumov,Maxim Samoylov,Maxime Lecanu,Maya Pavlova,Md Taha Bin Jawaid,Meghan Keneally,Melanie Kambadur,Meng Zhang,Mengchen Liu,Mengdi Lin,Mengjiao Wang,Mervyn Abraham,Miao Liu,Michael Au-Yeung,Michael Feldergraf,Michael Man,Michael Matheny,Michael Suo,Michael Tontchev,Michel Meyer,Michelle Ma,Mihir Patel,Mihir Sanjay Kale,Mik Vyatskov,Mikayla Alexander,Mike Andersland,Mike Clark,Mike Lewis,Mike Li,Mike Macey,Mike Macey,Mike Seltzer,Mikel Jimenez Fernandez,Mikhail Antonov,Mikhail Plekhanov,Milan Zhou,Min Si,Ming Qiao,Mingbo Ma,Mingjun Zhang,Mingyi Liang,Miquel Jubert Hermoso,Mirac Suzgun,Mirjam Skarica,Mitesh Kumar Singh,Mohammad Kabbani,Mohammad Rastegari,Mona Sarantakos,Monica Sim,Monika Gangapuram,Mor Moshe,Morrie Doulaty,Morvarid Metanat,Moya Chen,Mrinal Kumar,Munish Bansal,Murali Ramarao,Na Li,Nadav Azaria,Nahiyan Malik,Naman Goyal,Nancy Vargas Balderas,Nanshu Wang,Naoyuki Kanda,Natalia Gimelshein,Natalia Neverova,Nathan Aclander,Natt Sithiviraporn,Navneet Madhu Kumar,Ned Newton,Neeraj Bahl,Negar Ghorbani,Neil Patel,Neta-lee Golan,Nicholas Longenbaugh,Nick Egebo,Nikhil Johri,Nikhil Mehta,Nikhil Naik,Niko Moritz,Nikolay Bashlykov,Nikolay Bogoychev,Nikolay Pavlovich Laptev,Niladri Chatterji,Nile Jones,Nimish Shah,Ning Dong,Ning Li,Ning Li,Ning Zhang,Nishant Yadav,Noam Paz,Norman Cheng,Norman Cheng,Olaoluwa Adesanya,Oleg Repin,Oleksandr Maksymets,Omkar Salpekar,Omri Harosh,Onkar Pednekar,Onur Çelebi,Oran Gafni,Oren Edinger,Osama Hanna,Owais Khan Mohammed,Ozlem Kalinli,Paden Tomasello,Pankaj Singh,Paola Quevedo,Parag Jain,Paria Rashidinejad,Parker Tooley,Parth Parekh,Parth Thakkar,Parvin Taheri,Pasan Hapuarachchi,Pascal Kesseli,Patrick Alrassy,Paulo de Rezende Pinatti,Pavan Balaji,Pawan Sisodiya,Pedro Jose Ferreira Moreira,Pedro Rittner,Pedro Valenzuela,Peize Sun,Peizhao Zhang,Peng-Jen Chen,Pengchao Wang,Pengchuan Zhang,Pengwei Li,Petar Vasic,Peter Carras,Peter Ney,Peter Weng,Petru Dumea,Phil Hayes,Philip Woods,Pierre Andrews,Pierre Ménard,Ping-Hao Wu,Pingchuan Liu,Piotr Dollar,Plamen Dzhelepov,Polina Zvyagina,Posten A,Prabhav Agrawal,Pradhapan Rajendran,Pradyot Prakash,Prajjwal Bhargava,Pramono,Pranay Shah,Pranshu Dave,Prash Jain,Pratik Dubal,Praveen Gollakota,Praveen Krishnan,Pritish Yuvraj,Projjal Ghosh,Punit Singh Koura,Puxin Xu,Qi Qi,Qi Zhou,Qian Guan,Qian Sun,Qiang Liu,Qing He,Qinqing Zheng,Qirui Yang,Qizhen Guo,Quanzeng You,Quentin Carbonneaux,Quentin Carbonneaux,Quentin Duval,Quintin Fettes,Rachad Alao,Rachel Batish,Rachel Guo,Rachel Rodriguez,Radhika Bhargava,Rafael Asuncion,Raghotham Murthy,Rahul Dutta,Rahul Jha,Rahul Kindi,Rahul Mitra,Raj Ganapathy,Raj Shah,Rajarshi Das,Rajat Shrivastava,Rajesh Nishtala,Ramakant Shankar,Raman Shukhau,Ramon Calderer,Rangaprabhu Parthasarathy,Ranjan Subramanian,Raphael Bensadoun,Rares Bostan,Rashnil Chaturvedi,Ravi Agrawal,Ray Gao,Raymond Li,Rebecca Kogen,Ricardo Juan Palma Duran,Ricardo Silveira Cabral,Richard Lee,Richard Yuanzhe Pang,Riddhish Bhalodia,Riham Mansour,Rishabh Singh,Rishi Godugu,Ritun Patney,Rob Boyle,Robbie Goldfarb,Robert Caldwell,Robert Kuo,Roberta Raileanu,Robin Battey,Robin Sharma,Rochit Sapra,Rocky Wang,Rodolfo Granata,Rodrigo De Castro,Rodrigo Paim,Rohan Maheshwari,Rohan Varma,Rohit Girdhar,Rohit Patel,Roshan Sumbaly,Roy Sheaffer,Ruan Silva,Ruben Rodriguez Buchillon,Rui Hou,Ruiming Xie,Ruslan Mavlyutov,Ruslan Semenov,Rustam Dinov,Ruxiao Bao,Ryan Fox,Ryan Kilpatrick,Ryan Kwan,Ryan Lim,Ryan Smith,Saaketh Narayan,Sabrina Qiao,Sachin Mehta,Sachin Siby,Sagar Jain,Saghar Hosseini,Sagie Gur-Ari,Sahana Chennabasappa,Sahin Geyik,Sai Jayesh Bondu,Sai Mounika Chowdhary Nekkalapudi,Saif Hasan,Saisuke Okabayashi,Saketh Rambhatla,Salil Sawhney,Sam Dunster,Sam Zhao,Saman Keon,Samaneh Azadi,Sameet Sapra,Samuel Dooley,Samyak Datta,Sandeep Parab,Sang Michael Xie,Sanjay Singh,Sanyuan Chen,Sara Behn,Sara Khodeir,Sarah Shirazyan,Sargun Dhillon,Sarunya Pumma,Sasha Sidorov,Saskia Adaime,Saurabh Khanna,Sayem Wani,Scott Brenton,Sean Bell,Sean Kelly,Sean Koger,Sean Nunley,Sean Perry,Sebastian Caicedo,Sebastian Dahlgren,Sebastian Ruder,Seiji Yamamoto,Selam Mehretu,Selvan Sunitha Ravi,Sen Lyu,Senthil Chellapan,Serafeim Mellos,Sergey Edunov,Sergey Royt,Shaina Cohen,Shangfu Peng,Shannon Adams,Shaoliang Nie,Sharadh Ramaswamy,Sharan Narang,Shashank Pisupati,Shashi Gandham,Shaun Lim,Shaun Lindsay,Sheena Artrip,Shelly Sheynin,Shen Yan,Sheng Feng,Sheng Shen,Shengbao Zheng,Shenghao Lin,Shengjie Bi,Shengxin Cindy Zha,Shengye Wan,Shengyi Qian,Shengyong Cai,Shengzhi Shao,Shervin Shahidi,Shikai Li,Shimon Bernholtz,Shiqi Wang,Shishir G. Patil,Shiv Verma,Shiva Shankar P,Shiyang Chen,Sho Yaida,Shoubhik Debnath,Shreyas Siravara,Shruti Bhosale,Shuang Ma,Shun Zhang,Shuo Tang,Shuqiang Zhang,Shuyan Zhou,Sicong Che,Sidd Srinivisan,Siddharth Bhattacharya,Siddharth Patki,Sijia Chen,Sili Chen,Simon Vandenhende,Simone Merello,Sinong Wang,Sivan Barzily,Sixian Yi,Siyu Lin,SK Bong,Sky Yin,Sneha Agarwal,Sneha Agarwal,Soerian Lieve,Soji Sajuyigbe,Song Jiang,Songlin Li,Sonia Kim,Sopan Khosla,Soumi Maiti,Spencer Whitman,Sravya Popuri,Sreen Tallam,Srinivas Vaidyanathan,Srinivas Vaidyanathan,Sten Sootla,Stephane Collot,Stephanie Ding,Stephen Chen,Steven Cai,Suchin Gururangan,Sudarshan Govindaprasad,Sue Young,Suganthi Dewakar,Sujan Kumar Gonugondla,Sujeet Bhandari,Suman Gumudavelli,Suman Gumudavelli,Sumit Gupta,Summer Deng,Sungmin Cho,Suresh Ganapathy,Surjyendu Dhal,Susan Fedynak,Susana Contrera,Suyoun Kim,Sylvestre Rebuffi,Takshak Chahande,Tamar Herman,Tan Li,Tao Xu,Tara Fowler,Tarek Sheasha,Tarun Anand,Tarun Kalluri,Tarun Singh,Tatiana Shavrina,Ted Li,Teja Rao,Tejas Patil,Teng Li,Thach Bui,Thai Quach,Thamer Alharbash,Thanh Vinh Vo,Thawan Kooburat,Thilo Koehler,Thomas Georgiou,Thomas Scialom,Tian Ye,Tianhe Li,Tianjun Zhang,Tianyu Li,Tijmen Blankevoort,Timon Willi,Timothy Chou,Timothy Leung,TJ Lee,Todor Mihaylov,Tom Heatwole,Tong Xiao,Tony Cao,Tony Lee,Trang Le,Tristan Rice,Tsz Kei Serena Chan,Tuan Tran,Tudor Tiplea,Tyler Baumgartner,Uday Savagaonkar,Ujjwal Karn,Ulises Martinez Araiza,Umar Farooq,Uriel Cohen,Usman Sharif,Utkarsh Murarka,Van Phung,Varun Joginpalli,Varun Saravagi,Vasu Sharma,Vasudha Viswamurthy,Vedanuj Goswami,Vedika Seth,Venkat Ramesh,Venkat Ramesh,Vibhor Gupta,Victoria Montanez,Vidhya Natarajan,Vidya Sarma,Vignesh Ramanathan,Viktor Kerkez,Vinay Rao,Vincent Gonguet,Vincent Mauge,Virginie Do,Vish Vogeti,Vishrav Chaudhary,Viswesh Sankaran,Vítor Albiero,Vivek Miglani,Vivek Pai,Vlad Cojanu,Vlad Shubin,Vlad Tiberiu Mihailescu,Vladan Petrovic,Vladimir Ivanov,Vladislav Vorotilov,Vrushali Bhutada,Wai I Ng,Wei Cheng,Wei Sun,Wei Tu,Wei Wei,Wei Zhou,Wei-Ning Hsu,Weiwei Chu,Weizhe Yuan,Wenchen Wang,Wenjun Zhao,Wenwen Jiang,Wenyin Fu,Wenzhe Jiang,Whitney Meers,Will Constable,Will Wang,William R. Wong,Xavier Martinet,Xi Victoria Lin,Xi Yan,Xi Yin,Xian Li,Xianfeng Rui,Xianjun Yang,Xiaocheng Tang,Xiaodong Wang,Xiaofang Wang,Xiaolan Wang,Xiaoliang Dai,Xiaoliang Peng,Xiaopeng Li,Xiaozhu Meng,Xibei Zhang,Xide Xia,Xin Jin,xinbo Gao,Xinfeng Xie,Xingyi Zhou,Xu Ma,Xuan Ju,Xuanyi Zhao,Xubo Liu,Xuchao Jia,Xuedong Zhang,Xuefei Cao,Xuewei Wang,Xuewei Wu,Xunnan Xu,Xutai Ma,Xuyang Wang,Yan Cui,Yang Chen,Yang Li,Yang Shu,Yang Xia,Yanjun Chen,Yanjun Zhou,Yash Mehta,Yash Patel,Yash Tekena,Yashesh Gaur,Yasmine Babaei,Yaxuan Zhou,Ye Hu,Ye Qi,Yejin Lee,Yeming Wen,Yen-Cheng Liu,Yexin Bruce Wu,Yi Pan,Yi Yang,Yi-Hui Lin,Yifan Wang,Yifan Wu,Yifan Yang,Yifei Huang,Yiftah Ben Aharon,Yilin Yang,Yiling You,Ying Xu,Ying Zhang,Yingquan Yuan,Yingru Liu,Yingyi Ma,Yining Yang,Yiting Lu,Yonatan Komornik,Yongjie Lin,Yoni Goyhman,Yossi Moran Mamo,Youngjin Nam,Yu Wang,Yu Lu,Yu Zhao,Yu-Ho Hsieh,Yu-Jung Lo,Yuandong Tian,Yuanhan Zhang,Yuanhao Xiong,Yuanshun Yao,Yuchen Hao,Yuchen Zhang,Yuchuan Li,Yue Cao,Yue Yu,Yue Zhao,Yuhan Guo,Yuhao Wang,Yuheng Huang,Yujie Lu,Yujun Shi,Yulun Wang,Yun He,Yun Wang,Yundi Qian,Yunfan Wang,Yunhao Tang,Yuning Mao,Yunlu Li,Yuqi Dai,Yuriy Hulovatyy,Yushi Hu,Yuxuan Sun,Zach Rait,Zach Wentz,Zacharie Delpierre Coudert,Zachary Collins,Zahra Hankir,Zecheng He,Zeeshan Ahmed,Zeeshan Ahmed,Zef RosnBrick,Zhan Shu,Zhanna Rohalska,Zhaoduo Wen,Zhe Liu,Zhe Liu,Zhen Qiao,Zhenggang Xu,Zhengwen Zhou,Zhengxing Chen,Zhenyu Tang,Zhichen Wu,Zhicheng Ouyang,Zhihong Lei,Zhipeng Hong,Zhiping Xiu,Zhiwei Zhao,Zhong Meng,Zhou Jin,Zhouhao Zeng,Zichang Liu,Zihang Meng,Zihuan Qiao,Zinnia Zheng,Zixi Qi,Ziyi Luo,Zoe Foulkes Birkhead,Zoey Sun,Zohar Achdut
备注：15 pages

【50】Dynamical Systems Analysis Reveals Functional Regimes in Large Language Models
标题：动态系统分析揭示大型语言模型中的功能机制
链接：https://arxiv.org/abs/2601.11622

作者：Hassan Ugail,Newton Howard

【51】GRADE: Replacing Policy Gradients with Backpropagation for LLM Alignment
标题：等级：用反向传播取代政策要素以实现LLM一致
链接：https://arxiv.org/abs/2601.11574

作者：Lukas Abrie Nel

【52】Discrete Semantic States and Hamiltonian Dynamics in LLM Embedding Spaces
标题：LLM嵌入空间中的离散语义状态和Hamilton动力学
链接：https://arxiv.org/abs/2601.11572

作者：Timo Aukusti Laine
备注：23 pages, 5 figures

【53】MIMIC-RD: Can LLMs differentially diagnose rare diseases in real-world clinical settings?
标题：MIIC-RD：LLM能否在现实世界的临床环境中区分诊断罕见疾病？
链接：https://arxiv.org/abs/2601.11559

作者：Zilal Eiz AlDin,John Wu,Jeffrey Paul Fung,Jennifer King,Mya Watts,Lauren ONeill,Adam Richard Cross,Jimeng Sun
备注：5 pages

Graph相关(图学习|图神经网络|图优化等)(17篇)

【1】Riemannian Liquid Spatio-Temporal Graph Network
标题：Riemann液体时空图网络
链接：https://arxiv.org/abs/2601.14115

作者：Liangsi Lu,Jingchao Wang,Zhaorong Dai,Hanqian Liu,Yang Shi
备注：This paper has been accepted to The Web Conference 2026

【2】Autonomous Knowledge Graph Exploration with Adaptive Breadth-Depth Retrieval
标题：具有自适应广度-深度检索的自主知识图探索
链接：https://arxiv.org/abs/2601.13969

作者：Joaquín Polonuer,Lucas Vittor,Iñaki Arango,Ayush Noori,David A. Clifton,Luciano Del Corro,Marinka Zitnik

【3】Principled Latent Diffusion for Graphs via Laplacian Autoencoders
标题：通过拉普拉斯自动编码器实现图的原则潜在扩散
链接：https://arxiv.org/abs/2601.13780

作者：Antoine Siraudin,Christopher Morris
备注：Preprint, under review

【4】DRGW: Learning Disentangled Representations for Robust Graph Watermarking
标题：DRGW：学习解纠缠表示以实现鲁棒图水印
链接：https://arxiv.org/abs/2601.13569

作者：Jiasen Li,Yanwei Liu,Zhuoyi Shang,Xiaoyan Gu,Weiping Wang
备注：Published at The Web Conference 2026 (WWW '26)

【5】Graph Neural Networks are Heuristics
标题：图神经网络是启发式的
链接：https://arxiv.org/abs/2601.13465

作者：Yimeng Min,Carla P. Gomes

【6】Beyond Cosine Similarity: Taming Semantic Drift and Antonym Intrusion in a 15-Million Node Turkish Synonym Graph
标题：超越Cosine相似性：驯服1500万个节点土耳其同义词图中的语义漂移和反语入侵
链接：https://arxiv.org/abs/2601.13251

作者：Ebubekir Tosun,Mehmet Emin Buldur,Özay Ezerceli,Mahmoud ElHussieni

【7】HT-GNN: Hyper-Temporal Graph Neural Network for Customer Lifetime Value Prediction in Baidu Ads
标题：HT-GNN：超时态图神经网络在百度广告中用于客户终身价值预测
链接：https://arxiv.org/abs/2601.13013

作者：Xiaohui Zhao,Xinjian Zhao,Jiahui Zhang,Guoyu Liu,Houzhi Wang,Shu Wu

【8】OFA-MAS: One-for-All Multi-Agent System Topology Design based on Mixture-of-Experts Graph Generative Models
标题：OFA-MAS：基于混合专家图生成模型的一对所有多智能体系统布局设计
链接：https://arxiv.org/abs/2601.12996

作者：Shiyuan Li,Yixin Liu,Yu Zheng,Mei Li,Quoc Viet Hung Nguyen,Shirui Pan
备注：Accepted by WWW 2026

【9】Deep Temporal Graph Clustering: A Comprehensive Benchmark and Datasets
标题：深度时间图集群：全面的基准和数据集
链接：https://arxiv.org/abs/2601.12903

作者：Meng Liu,Ke Liang,Siwei Wang,Xingchen Hu,Sihang Zhou,Xinwang Liu

【10】Eddy-Resolving Global Ocean Forecasting with Multi-Scale Graph Neural Networks
标题：利用多尺度图神经网络进行涡解全球海洋预测
链接：https://arxiv.org/abs/2601.12775

作者：Yuta Hirabayashi,Daisuke Matusoka,Konobu Kimura

【11】A Boolean Function-Theoretic Framework for Expressivity in GNNs with Applications to Fair Graph Mining
标题：GNN表达性的布尔函数理论框架及其在公平图挖掘中的应用
链接：https://arxiv.org/abs/2601.12751

作者：Manjish Pal

【12】A Graph Prompt Fine-Tuning Method for WSN Spatio-Temporal Correlation Anomaly Detection
标题：一种用于无线传感器网络时空相关异常检测的图提示微调方法
链接：https://arxiv.org/abs/2601.12745

作者：Miao Ye,Jing Cui,Yuan huang,Qian He,Yong Wang,Jiwen Zhang

【13】Decentralized Learning Strategies for Estimation Error Minimization with Graph Neural Networks
标题：利用图神经网络实现估计误差最小化的分散学习策略
链接：https://arxiv.org/abs/2601.12662

作者：Xingran Chen,Navid NaderiAlizadeh,Alejandro Ribeiro,Shirin Saeedi Bidokhti

【14】Graph Attention Networks with Physical Constraints for Anomaly Detection
标题：具有物理约束的图注意力网络用于异常检测
链接：https://arxiv.org/abs/2601.12426

作者：Mohammadhossein Homaei,Iman Khazrak,Ruben Molano,Andres Caro,Mar Avila
备注：7 Pages, 4 Figures, 5 Tables

【15】Learning Audio-Visual Embeddings with Inferred Latent Interaction Graphs
标题：使用推断潜在交互图学习视听嵌入
链接：https://arxiv.org/abs/2601.11995

作者：Donghuo Zeng,Hao Niu,Yanan Wang,Masato Taya
备注：16 pages, 5 figures, 2 tables

【16】Data-centric Prompt Tuning for Dynamic Graphs
标题：以数据为中心的动态图形提示调优
链接：https://arxiv.org/abs/2601.11954

作者：Yufei Peng,Cheng Yang,Zhengjie Fan,Chuan Shi
备注：CIKM 2025

【17】DeepEvidence: Empowering Biomedical Discovery with Deep Knowledge Graph Research
标题：DeepEvidence：通过深度知识图谱研究增强生物医学发现能力
链接：https://arxiv.org/abs/2601.11560

作者：Zifeng Wang,Zheng Chen,Ziwei Yang,Xuan Wang,Qiao Jin,Yifan Peng,Zhiyong Lu,Jimeng Sun

Transformer(22篇)

【1】A model of errors in transformers
标题：Transformer误差模型
链接：https://arxiv.org/abs/2601.14175

作者：Suvrat Raju,Praneeth Netrapalli
备注：8+17pages

【2】Learning to Explain: Supervised Token Attribution from Transformer Attention Patterns
标题：学习解释：来自Transformer注意力模式的监督代币归因
链接：https://arxiv.org/abs/2601.14112

作者：George Mihaila

【3】Two-Stream temporal transformer for video action classification
标题：用于视频动作分类的两流时间Transformer
链接：https://arxiv.org/abs/2601.14086

作者：Nattapong Kurpukdee,Adrian G. Bors

【4】HiT: History-Injection Transformers for Onboard Continuous Flood Change Detection
标题：HiT：用于机载连续洪水变化检测的历史注入Transformer
链接：https://arxiv.org/abs/2601.13751

作者：Daniel Kyselica,Jonáš Herec,Oliver Kutis,Rado Pitoňák
备注：19 pages, 9 figures, submitted to conference

【5】Neural Organ Transplantation (NOT): Checkpoint-Based Modular Adaptation for Transformer Models
标题：神经器官移植（NOT）：基于检查点的Transformer模型模块化适应
链接：https://arxiv.org/abs/2601.13580

作者：Ahmad Al-Zuraiqi
备注：27 pages, 8 figures, 16 tables. Decoder-only transformers (124M-20B parameters). Complete experimental results and reproducibility details in appendices. Code and checkpoints: https://github.com/zuraiqi/neural-organ-transplant

【6】A Learnable Wavelet Transformer for Long-Short Equity Trading and Risk-Adjusted Return Optimization
标题：用于长短股票交易和风险调整回报优化的可学习的子波Transformer
链接：https://arxiv.org/abs/2601.13435

作者：Shuozhe Li,Du Cheng,Leqi Liu

【7】Early Prediction of Type 2 Diabetes Using Multimodal data and Tabular Transformers
标题：使用多峰数据和表格变形器早期预测2型糖尿病
链接：https://arxiv.org/abs/2601.12981

作者：Sulaiman Khan,Md. Rafiul Biswas,Zubair Shah
备注：08 pages, 06 figures, accepted for publication in FLLM2025

【8】BlocksecRT-DETR: Decentralized Privacy-Preserving and Token-Efficient Federated Transformer Learning for Secure Real-Time Object Detection in ITS
标题：BlocksecRT-DETR：去中心化隐私保护和令牌高效的联邦Transformer学习，用于ITS中的安全实时对象检测
链接：https://arxiv.org/abs/2601.12693

作者：Mohoshin Ara Tahera,Sabbir Rahman,Shuvalaxmi Dass,Sharif Ullah,Mahmoud Abouyessef

【9】Patch-Level Tokenization with CNN Encoders and Attention for Improved Transformer Time-Series Forecasting
标题：使用CNN编码器和注意力的补丁级令牌化改进Transformer时间序列预测
链接：https://arxiv.org/abs/2601.12467

作者：Saurish Nagrath
备注：6 pages, 2 figures, 3 tables

【10】LiQSS: Post-Transformer Linear Quantum-Inspired State-Space Tensor Networks for Real-Time 6G
标题：LiQSS：用于实时6G的后Transformer线性量子启发状态空间张量网络
链接：https://arxiv.org/abs/2601.12375

作者：Farhad Rezazadeh,Hatim Chergui,Mehdi Bennis,Houbing Song,Lingjia Liu,Dusit Niyato,Merouane Debbah
备注：14 pages, 4 figures, 5 tables

【11】HCFT: Hierarchical Convolutional Fusion Transformer for EEG Decoding
标题：HCFT：用于脑电解码的分层卷积融合Transformer
链接：https://arxiv.org/abs/2601.12279

作者：Haodong Zhang,Jiapeng Zhu,Yitong Chen,Hongqi Li
备注：Submitted to IEEE Journals

【12】Neural Isomorphic Fields: A Transformer-based Algebraic Numerical Embedding
标题：神经同质场：基于变换器的代数数字嵌入
链接：https://arxiv.org/abs/2601.12095

作者：Hamidreza Sadeghi,Saeedeh Momtazi,Reza Safabakhsh

【13】TF-CoDiT: Conditional Time Series Synthesis with Diffusion Transformers for Treasury Futures
标题：TF-CoDiT：国债期货使用扩散变换器的条件时间序列合成
链接：https://arxiv.org/abs/2601.11880

作者：Yingxiao Zhang,Jiaxin Duan,Junfu Zhang,Ke Feng

【14】Cascaded Transformer for Robust and Scalable SLA Decomposition via Amortized Optimization
标题：通过摊销优化实现稳健且可扩展的SLA分解的级联Transformer
链接：https://arxiv.org/abs/2601.11859

作者：Cyril Shih-Huan Hsu

【15】Mixture of Distributions Matters: Dynamic Sparse Attention for Efficient Video Diffusion Transformers
标题：分布的混合很重要：高效视频扩散变形机的动态稀疏注意力
链接：https://arxiv.org/abs/2601.11641

作者：Yuxi Liu,Yipeng Hu,Zekun Zhang,Kunze Jiang,Kun Yuan

【16】NoiseFormer -- Noise Diffused Symmetric Attention Transformer
标题：NoiseFormer --噪音扩散对称注意力Transformer
链接：https://arxiv.org/abs/2601.11619

作者：Phani Kumar,Nyshadham,Jyothendra Varma,Polisetty V R K,Aditya Rathore

【17】Geometric Attention: A Regime-Explicit Operator Semantics for Transformer Attention
标题：几何注意力：Transformer注意力的区显式运算符语义
链接：https://arxiv.org/abs/2601.11618

作者：Luis Rosario Freytes
备注：57 pages

【18】Multi-modal MRI-Based Alzheimer's Disease Diagnosis with Transformer-based Image Synthesis and Transfer Learning
标题：利用基于变换器的图像合成和传输学习进行基于多模式MRI的阿尔茨海默病诊断
链接：https://arxiv.org/abs/2601.11614

作者：Jason Qiu
备注：19 pages, 10 figures

【19】Domain-Specific Self-Supervised Pre-training for Agricultural Disease Classification: A Hierarchical Vision Transformer Study
标题：农业疾病分类的特定领域自我监督预训练：分层视觉Transformer研究
链接：https://arxiv.org/abs/2601.11612

作者：Arnav S. Sonavane
备注：11 pages, 4 figures, 9 tables

【20】Forecasting Continuum Intensity for Solar Active Region Emergence Prediction using Transformers
标题：使用Transformer预测太阳活动区出现的连续体强度
链接：https://arxiv.org/abs/2601.13144

作者：Jonas Tirona,Sarang Patil,Spiridon Kasapis,Eren Dogan,John Stefan,Irina N. Kitiashvili,Alexander G. Kosovichev,Mengjia Xu
备注：30 pages, 7 figures, submitted to JGR: Machine Learning and Computation

【21】A Mixture of Experts Vision Transformer for High-Fidelity Surface Code Decoding
标题：用于高保真表面代码解码的混合专家视觉Transformer
链接：https://arxiv.org/abs/2601.12483

作者：Hoang Viet Nguyen,Manh Hung Nguyen,Hoang Ta,Van Khu Vu,Yeow Meng Chee
备注：16 pages, 7 figures

【22】Nonlinear Dynamic Factor Analysis With a Transformer Network
标题：Transformer网络的非线性动态因素分析
链接：https://arxiv.org/abs/2601.12039

作者：Oliver Snellman
备注：Working paper. 88 pages, 57 figures, 14 tables. Earlier versions circulated as "Nowcasting with a Transformer Network" (first version: 26 Oct 2024)

GAN|对抗|攻击|生成相关(17篇)

【1】SecureSplit: Mitigating Backdoor Attacks in Split Learning
标题：SecureSplit：缓解Split学习中的后门攻击
链接：https://arxiv.org/abs/2601.14054

作者：Zhihao Dou,Dongfei Cui,Weida Wang,Anjun Gao,Yueyang Quan,Mengyao Ma,Viet Vo,Guangdong Bai,Zhuqing Liu,Minghong Fang
备注：To appear in The Web Conference 2026

【2】PAC-Private Responses with Adversarial Composition
标题：PAC-具有对抗性成分的私人回应
链接：https://arxiv.org/abs/2601.14033

作者：Xiaochen Zhu,Mayuri Sridhar,Srinivas Devadas
备注：16 pages, 3 figures

【3】Asymmetric regularization mechanism for GAN training with Variational Inequalities
标题：具有变分不等式的GAN训练的非对称正规化机制
链接：https://arxiv.org/abs/2601.13920

作者：Spyridon C. Giagtzoglou,Mark H. M. Winands,Barbara Franci
备注：6 pages, 3 figures, conference

【4】MN-TSG:Continuous Time Series Generation with Irregular Observations
标题：MN-TSG：使用不规则观察生成连续时间序列
链接：https://arxiv.org/abs/2601.13534

作者：Xu Zhang,Junwei Deng,Chang Xu,Hao Li,Jiang Bian
备注：34 pages

【5】A Hybrid Protocol for Large-Scale Semantic Dataset Generation in Low-Resource Languages: The Turkish Semantic Relations Corpus
标题：低资源语言中大规模语义数据集生成的混合协议：土耳其语语义关系语料库
链接：https://arxiv.org/abs/2601.13253

作者：Ebubekir Tosun,Mehmet Emin Buldur,Özay Ezerceli,Mahmoud ElHussieni

【6】Diffusion-Driven Synthetic Tabular Data Generation for Enhanced DoS/DDoS Attack Classification
标题：扩散驱动的合成表格数据生成，用于增强的NOS/DDOS攻击分类
链接：https://arxiv.org/abs/2601.13197

作者：Aravind B,Anirud R. S.,Sai Surya Teja N,Bala Subrahmanya Sriranga Navaneeth A,Karthika R,Mohankumar N
备注：7 pages, 8 figures, 2025 International Conference on Signal Processing, Computation, Electronics, Power and Telecommunication (IConSCEPT), National Institute of Technology, Puducherry, India

【7】LAViG-FLOW: Latent Autoregressive Video Generation for Fluid Flow Simulations
标题：LAViG-FLOW：用于流体流动模拟的潜在自回归视频生成
链接：https://arxiv.org/abs/2601.13190

作者：Vittoria De Pellegrini,Tariq Alkhalifah

【8】NeuroShield: A Neuro-Symbolic Framework for Adversarial Robustness
标题：NeuShield：对抗稳健性的神经符号框架
链接：https://arxiv.org/abs/2601.13162

作者：Ali Shafiee Sarvestani,Jason Schmidt,Arman Roohi

【9】Generating Cyclic Conformers with Flow Matching in Cremer-Pople Coordinates
标题：利用Cremer-Pople坐标系中的流匹配生成循环适形体
链接：https://arxiv.org/abs/2601.12859

作者：Luca Schaufelberger,Aline Hartgers,Kjell Jorner

【10】Joint Source-Channel-Generation Coding: From Distortion-oriented Reconstruction to Semantic-consistent Generation
标题：联合信源-通道-生成编码：从面向失真的重建到语义一致的生成
链接：https://arxiv.org/abs/2601.12808

作者：Tong Wu,Zhiyong Chen,Guo Lu,Li Song,Feng Yang,Meixia Tao,Wenjun Zhang
备注：submitted to IEEE ISIT 2026

【11】Towards Robust Universal Perturbation Attacks: A Float-Coded, Penalty-Driven Evolutionary Approach
标题：走向稳健的普遍扰动攻击：浮动编码、惩罚驱动的进化方法
链接：https://arxiv.org/abs/2601.12624

作者：Shiqi Wang,Mahdi Khosravy,Neeraj Gupta,Olaf Witkowski

【12】Beyond the Dirac Delta: Mitigating Diversity Collapse in Reinforcement Fine-Tuning for Versatile Image Generation
标题：超越狄拉克三角洲：缓解多功能图像生成的强化微调中的多样性崩溃
链接：https://arxiv.org/abs/2601.12401

作者：Jinmei Liu,Haoru Li,Zhenhong Sun,Chaofeng Chen,Yatao Bian,Bo Wang,Daoyi Dong,Chunlin Chen,Zhi Wang

【13】DevBench: A Realistic, Developer-Informed Benchmark for Code Generation Models
标题：DevBench：一个现实的、受操作员知情的代码生成模型基准
链接：https://arxiv.org/abs/2601.11895

作者：Pareesa Ameneh Golnari,Adarsh Kumarappan,Wen Wen,Xiaoyu Liu,Gabriel Ryan,Yuting Sun,Shengyu Fu,Elsie Nallipogu

【14】Semantic Differentiation for Tackling Challenges in Watermarking Low-Entropy Constrained Generation Outputs
标题：语义差异以应对水印低熵约束发电输出的挑战
链接：https://arxiv.org/abs/2601.11629

作者：Nghia T. Le,Alan Ritter,Kartik Goyal
备注：18 pages, 4 figures

【15】Generative Adversarial Networks for Resource State Generation
标题：用于资源状态生成的生成对抗网络
链接：https://arxiv.org/abs/2601.13708

作者：Shahbaz Shaik,Sourav Chatterjee,Sayantan Pramanik,Indranil Chakrabarty

【16】Beyond Visual Realism: Toward Reliable Financial Time Series Generation
标题：超越视觉现实主义：迈向可靠的金融时间序列生成
链接：https://arxiv.org/abs/2601.12990

作者：Fan Zhang,Jiabin Luo,Zheng Zhang,Shuanghong Huang,Zhipeng Liu,Yu Chen
备注：Accepted by ICASSP 2026

【17】Adversarial Drift-Aware Predictive Transfer: Toward Durable Clinical AI
标题：对抗性漂移感知预测转移：迈向持久临床人工智能
链接：https://arxiv.org/abs/2601.11860

作者：Xin Xiong,Zijian Guo,Haobo Zhu,Chuan Hong,Jordan W Smoller,Tianxi Cai,Molei Liu

半/弱/无/有监督|不确定性|主动学习(14篇)

【1】Unsupervised Video Class-Incremental Learning via Deep Embedded Clustering Management
标题：无监督视频类-通过深度嵌入式聚类管理进行增量学习
链接：https://arxiv.org/abs/2601.14069

作者：Nattapong Kurpukdee,Adrian G. Bors

【2】Group-Invariant Unsupervised Skill Discovery: Symmetry-aware Skill Representations for Generalizable Behavior
标题：群体不变的无监督技能发现：可概括行为的对称性感知技能表示
链接：https://arxiv.org/abs/2601.14000

作者：Junwoo Chang,Joseph Park,Roberto Horowitz,Jongmin Lee,Jongeun Choi
备注：14 pages, 6 figures

【3】RL-BioAug: Label-Efficient Reinforcement Learning for Self-Supervised EEG Representation Learning
标题：RL-BioAug：用于自我监督脑电表示学习的标签高效强化学习
链接：https://arxiv.org/abs/2601.13964

作者：Cheol-Hui Lee,Hwa-Yeon Lee,Dong-Joo Kim

【4】Who Should Have Surgery? A Comparative Study of GenAI vs Supervised ML for CRS Surgical Outcome Prediction
标题：谁应该接受手术？GenAI与监督ML预测CRS手术结局的比较研究
链接：https://arxiv.org/abs/2601.13710

作者：Sayeed Shafayet Chowdhury,Snehasis Mukhopadhyay,Shiaofen Fang,Vijay R. Ramakrishnan

【5】Uncertainty-Aware Gradient Signal-to-Noise Data Selection for Instruction Tuning
标题：用于指令调谐的不确定性感知梯度信噪比数据选择
链接：https://arxiv.org/abs/2601.13697

作者：Zhihang Yuan,Chengyu Yue,Long Huang,Litu Ou,Lei Shi
备注：Preprint

【6】Multi-level Monte Carlo Dropout for Efficient Uncertainty Quantification
标题：多级别蒙特卡洛退出，实现高效的不确定性量化
链接：https://arxiv.org/abs/2601.13272

作者：Aaron Pim,Tristan Pryer
备注：26 pages, 11 figures

【7】Supervised Learning for the (s,S) Inventory Model with General Interarrival Demands and General Lead Times
标题：具有一般到达间隔需求和一般交货时间的（s，S）库存模型的监督学习
链接：https://arxiv.org/abs/2601.12900

作者：Eliran Sherzer,Yonit Barron

【8】Constraint-Aware Neurosymbolic Uncertainty Quantification with Bayesian Deep Learning for Scientific Discovery
标题：基于贝叶斯深度学习的约束感知神经符号不确定性量化
链接：https://arxiv.org/abs/2601.12442

作者：Shahnawaz Alam,Mohammed Mudassir Uddin,Mohammed Kaif Pasha

【9】Task-tailored Pre-processing: Fair Downstream Supervised Learning
标题：任务定制的预处理：公平的下游监督学习
链接：https://arxiv.org/abs/2601.11897

作者：Jinwon Sohn,Guang Lin,Qifan Song

【10】A Confidence-Variance Theory for Pseudo-Label Selection in Semi-Supervised Learning
标题：半监督学习中伪标签选择的置信方差理论
链接：https://arxiv.org/abs/2601.11670

作者：Jinshi Liu,Pan Liu

【11】Co-Initialization of Control Filter and Secondary Path via Meta-Learning for Active Noise Control
标题：通过元学习协调控制过滤器和辅助路径以实现主动噪音控制
链接：https://arxiv.org/abs/2601.13849

作者：Ziyi Yang,Li Rao,Zhengding Luo,Dongyuan Shi,Qirui Huang,Woon-Seng Gan

【12】SolARED: Solar Active Region Emergence Dataset for Machine Learning Aided Predictions
标题：SolARED：用于机器学习辅助预测的太阳活跃区出现数据集
链接：https://arxiv.org/abs/2601.13145

作者：Spiridon Kasapis,Eren Dogan,Irina N. Kitiashvili,Alexander G. Kosovichev,John T. Stefan,Jake D. Butler,Jonas Tirona,Sarang Patil,Mengjia Xu
备注：15 pages, 6 figures, submitted to the Springer Nature - Solar Physics Journal

【13】Gradient-based Active Learning with Gaussian Processes for Global Sensitivity Analysis
标题：基于对象的主动学习和高斯过程进行全局敏感性分析
链接：https://arxiv.org/abs/2601.11790

作者：Guerlain Lambert,Céline Helbert,Claire Lauvernet

【14】Lightweight Self-Supervised Detection of Fundamental Frequency and Accurate Probability of Voicing in Monophonic Music
标题：单音音乐中基本频率和准确发声概率的轻量级自监督检测
链接：https://arxiv.org/abs/2601.11768

作者：Venkat Suprabath Bitra,Homayoon Beigi
备注：12 pages, 6 figures, 3 tables, and an appendix, Accepted for publication at ICPRAM 2026 in Marbella, Spain, on March 2, 2026

迁移|Zero/Few/One-Shot|自适应(11篇)

【1】AdaNODEs: Test Time Adaptation for Time Series Forecasting Using Neural ODEs
标题：AdaNODEs：使用神经ODE进行时间序列预测的测试时间自适应
链接：https://arxiv.org/abs/2601.12893

作者：Ting Dang,Soumyajit Chatterjee,Hong Jia,Yu Wu,Flora Salim,Fahim Kawsar
备注：Accepted by ICASSP 2026

【2】Adaptive Multi-Scale Correlation Meta-Network for Few-Shot Remote Sensing Image Classification
标题：用于Few-Shot遥感图像分类的自适应多尺度相关元网络
链接：https://arxiv.org/abs/2601.12308

作者：Anurag Kaushish,Ayan Sar,Sampurna Roy,Sudeshna Chakraborty,Prashant Trivedi,Tanupriya Choudhury,Kanav Gupta
备注：Accepted in IEEE ICASSP 2026

【3】TimeGMM: Single-Pass Probabilistic Forecasting via Adaptive Gaussian Mixture Models with Reversible Normalization
标题：TimeGMM：基于可逆归一化的自适应高斯混合模型的单遍概率预测
链接：https://arxiv.org/abs/2601.12288

作者：Lei Liu,Tengyuan Liu,Hongwei Zhao,Jiahui Huang,Ruibo Guo,Bin Li

【4】Wavelet-Aware Anomaly Detection in Multi-Channel User Logs via Deviation Modulation and Resolution-Adaptive Attention
标题：通过偏差调制和分辨率自适应注意力在多通道用户搜索中进行子波感知异常检测
链接：https://arxiv.org/abs/2601.12231

作者：Kaichuan Kong,Dongjie Liu,Xiaobo Jin,Shijie Xu,Guanggang Geng
备注：Accepted by ICASSP 2026. Copyright 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

【5】PTL-PINNs: Perturbation-Guided Transfer Learning with Physics- Informed Neural Networks for Nonlinear Systems
标题：PTL-PINNs：用于非线性系统的物理信息神经网络的扰动引导迁移学习
链接：https://arxiv.org/abs/2601.12093

作者：Duarte Alexandrino,Ben Moseley,Pavlos Protopapas
备注：51 pages, 14 figures, 7 tables

【6】One-Shot Price Forecasting with Covariate-Guided Experts under Privacy Constraints
标题：隐私约束下协变量引导专家的一次性价格预测
链接：https://arxiv.org/abs/2601.11977

作者：Ren He,Yinliang Xu,Jinfeng Wang,Jeremy Watson,Jian Song

【7】Harmonica: A Self-Adaptation Exemplar for Sustainable MLOps
标题：Harmonica：可持续MLOps的自适应典范
链接：https://arxiv.org/abs/2601.11926

作者：Ananya Halgatti,Shaunak Biswas,Hiya Bhatt,Srinivasan Rakhunathan,Karthik Vaidhyanathan
备注：This paper has been accepted to SEAMS 2026 Artifact Track

【8】IPEC: Test-Time Incremental Prototype Enhancement Classifier for Few-Shot Learning
标题：ITEC：用于Few-Shot学习的测试时增量原型增强分类器
链接：https://arxiv.org/abs/2601.11669

作者：Wenwen Liao,Hang Ruan,Jianbo Yu,Xiaofeng Yang,Qingchao Jiang,Xuefeng Yan

【9】AdaFRUGAL: Adaptive Memory-Efficient Training with Dynamic Control
标题：AdaFRUGAL：具有动态控制的自适应记忆高效训练
链接：https://arxiv.org/abs/2601.11568

作者：Quang-Hung Bui,Anh Son Ta

【10】Energy-Efficient Prediction in Textile Manufacturing: Enhancing Accuracy and Data Efficiency With Ensemble Deep Transfer Learning
标题：纺织制造业的节能预测：通过融合深度迁移学习提高准确性和数据效率
链接：https://arxiv.org/abs/2601.12663

作者：Yan-Chen Chen,Wei-Yu Chiu,Qun-Yu Wang,Jing-Wei Chen,Hao-Ting Zhao
备注：26 pages, 11 figures

【11】Adaptive Rotary Steering with Joint Autoregression for Robust Extraction of Closely Moving Speakers in Dynamic Scenarios
标题：具有联合自回归的自适应旋转转向，用于动态场景中近距离移动的扬声器的鲁棒提取
链接：https://arxiv.org/abs/2601.12345

作者：Jakob Kienegger,Timo Gerkmann
备注：Accepted at IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2026

强化学习(15篇)

【1】Jet-RL: Enabling On-Policy FP8 Reinforcement Learning with Unified Training and Rollout Precision Flow
标题：Jet-RL：通过统一训练和推出精确流程实现按政策FP 8强化学习
链接：https://arxiv.org/abs/2601.14243

作者：Haocheng Xi,Charlie Ruan,Peiyuan Liao,Yujun Lin,Han Cai,Yilong Zhao,Shuo Yang,Kurt Keutzer,Song Han,Ligeng Zhu
备注：11 pages, 6 figures, 4 tables

【2】Spatiotemporal Wildfire Prediction and Reinforcement Learning for Helitack Suppression
标题：用于直升机抑制的时空野火预测和强化学习
链接：https://arxiv.org/abs/2601.14238

作者：Shaurya Mathur,Shreyas Bellary Manjunath,Nitin Kulkarni,Alina Vereshchaka
备注：6 pages, 5 figures (two of them in tables), Conference: IEEE International Conference on Machine Learning and Applications 2025 (ICMLA 2025): https://www.icmla-conference.org/icmla25/

【3】KAGE-Bench: Fast Known-Axis Visual Generalization Evaluation for Reinforcement Learning
标题：KAGE-Bench：强化学习的快速已知轴视觉概括评估
链接：https://arxiv.org/abs/2601.14232

作者：Egor Cherepanov,Daniil Zelezetsky,Alexey K. Kovalev,Aleksandr I. Panov
备注：38 pages, 44 figures, 3 tables

【4】Attention-Based Offline Reinforcement Learning and Clustering for Interpretable Sepsis Treatment
标题：基于注意力的离线强化学习和集群用于可解释败血症治疗
链接：https://arxiv.org/abs/2601.14228

作者：Punit Kumar,Vaibhav Saran,Divyesh Patel,Nitin Kulkarni,Alina Vereshchaka
备注：8 pages, 6 figures, Conference: IEEE International Conference on Machine Learning and Applications 2025 (ICMLA 2025): https://www.icmla-conference.org/icmla25/

【5】Optimizing Energy and Data Collection in UAV-aided IoT Networks using Attention-based Multi-Objective Reinforcement Learning
标题：使用基于注意力的多目标强化学习优化无人机辅助物联网网络中的能量和数据收集
链接：https://arxiv.org/abs/2601.14092

作者：Babacar Toure,Dimitrios Tsilimantos,Omid Esrafilian,Marios Kountouris

【6】Reinforcement Learning for Opportunistic Routing in Software-Defined LEO-Terrestrial Systems
标题：软件定义LEO-地面系统中启发式路由的强化学习
链接：https://arxiv.org/abs/2601.13662

作者：Sivaram Krishnan,Zhouyou Gu,Jihong Park,Sung-Min Oh,Jinho Choi

【7】Communication-Free Collective Navigation for a Swarm of UAVs via LiDAR-Based Deep Reinforcement Learning
标题：通过基于LiDART的深度强化学习为大量无人机提供无通信的集体导航
链接：https://arxiv.org/abs/2601.13657

作者：Myong-Yol Choi,Hankyoul Ko,Hanse Cho,Changseung Kim,Seunghwan Kim,Jaemin Seo,Hyondong Oh

【8】Communication Methods in Multi-Agent Reinforcement Learning
标题：多智能体强化学习中的通信方法
链接：https://arxiv.org/abs/2601.12886

作者：Christoph Wittner
备注：12 pages, 2 figures

【9】Optimal Power Allocation and Sub-Optimal Channel Assignment for Downlink NOMA Systems Using Deep Reinforcement Learning
标题：使用深度强化学习的下行NOMA系统的最佳功率分配和次优通道分配
链接：https://arxiv.org/abs/2601.12242

作者：WooSeok Kim,Jeonghoon Lee,Sangho Kim,Taesun An,WonMin Lee,Dowon Kim,Kyungseop Shin

【10】Speculative Sampling with Reinforcement Learning
标题：基于强化学习的推测采样
链接：https://arxiv.org/abs/2601.12212

作者：Chenan Wang,Daniel H. Shi,Haipeng Chen
备注：Accepted to AAAI 2026

【11】Extreme Value Policy Optimization for Safe Reinforcement Learning
标题：安全强化学习的极端值策略优化
链接：https://arxiv.org/abs/2601.12008

作者：Shiqing Gao,Yihang Zhou,Shuai Shao,Haoyu Luo,Yiheng Bing,Jiaxin Ding,Luoyi Fu,Xinbing Wang
备注：Published in the 42nd International Conference on Machine Learning (ICML 2025)

【12】Controlling Underestimation Bias in Constrained Reinforcement Learning for Safe Exploration
标题：安全探索中约束强化学习的低估偏差控制
链接：https://arxiv.org/abs/2601.11953

作者：Shiqing Gao,Jiaxin Ding,Luoyi Fu,Xinbing Wang
备注：Published in the 42nd International Conference on Machine Learning (ICML 2025, Oral)

【13】Reinforcement Learning for Dynamic Workflow Optimization in CI/CD Pipelines
标题：CI/CD管道中动态工作流程优化的强化学习
链接：https://arxiv.org/abs/2601.11647

作者：Aniket Abhishek Soni,Milan Parikh,Rashi Nimesh Kumar Dhenia,Jubin Abhishek Soni,Ayush Raj Jha,Sneja Mitinbhai Shah
备注：Accepted and presented at CICN 2025 (International Conference on Computational Intelligence and Communication Networks). 7 pages, 5 figures

【14】Hindsight Preference Replay Improves Preference-Conditioned Multi-Objective Reinforcement Learning
标题：事后诸葛亮偏好重播改进了偏好条件多目标强化学习
链接：https://arxiv.org/abs/2601.11604

作者：Jonaid Shianifar,Michael Schukat,Karl Mason

【15】Sample Complexity of Average-Reward Q-Learning: From Single-agent to Federated Reinforcement Learning
标题：平均奖励Q学习的样本复杂性：从单智能体到联邦强化学习
链接：https://arxiv.org/abs/2601.13642

作者：Yuchen Jiao,Jiin Woo,Gen Li,Gauri Joshi,Yuejie Chi

元学习(1篇)

【1】Recursive Meta-Distillation: An Axiomatic Framework for Iterative Knowledge Refinement
标题：回归元蒸馏：迭代知识精炼的公理框架
链接：https://arxiv.org/abs/2601.13100

作者：Aaron R. Flouro,Shawn P. Chadwick

符号|符号学习(2篇)

【1】Breaking the Data Barrier in Learning Symbolic Computation: A Case Study on Variable Ordering Suggestion for Cylindrical Algebraic Decomposition
标题：打破学习符号计算的数据障碍：圆柱形代数分解的变量排序建议案例研究
链接：https://arxiv.org/abs/2601.13731

作者：Rui-Juan Jing,Yuegang Zhao,Changbo Chen

【2】CSyMR: Benchmarking Compositional Symbolic Muisc Reasoning With MIR Tool Integration
标题：CSyMR：通过MIR工具集成对合成符号Muisc推理进行基准测试
链接：https://arxiv.org/abs/2601.11556

作者：Boyang Wang,Yash Vishe,Xin Xu,Zachary Novack,Julian McAuley,Junda Wu

分层学习(1篇)

【1】Device Association and Resource Allocation for Hierarchical Split Federated Learning in Space-Air-Ground Integrated Network
标题：空地一体化网络分层分离联邦学习的设备关联和资源分配
链接：https://arxiv.org/abs/2601.13817

作者：Haitao Zhao,Xiaoyu Tang,Bo Xu,Jinlong Sun,Linghao Zhang

医学相关(11篇)

【1】From 100,000+ images to winning the first brain MRI foundation model challenges: Sharing lessons and models
标题：从100，000多张图像到赢得首个脑部MRI基金会模型挑战：分享教训和模型
链接：https://arxiv.org/abs/2601.13166

作者：Pedro M. Gordaliza,Jaume Banus,Benoît Gérin,Maxence Wynen,Nataliia Molchanova,Jonas Richiardi,Meritxell Bach Cuadra
备注：Work presented at the SSL3D Challenge (1st place, ResEnc-L track) and FOMO Challenge (1st place, Methods track) on Brain MRI Foundation Models at MICCAI 2025

【2】Enhancing Generalization in Sickle Cell Disease Diagnosis through Ensemble Methods and Feature Importance Analysis
标题：通过整合方法和特征重要性分析增强镰状细胞病诊断的概括性
链接：https://arxiv.org/abs/2601.13021

作者：Nataša Petrović,Gabriel Moyà-Alcover,Antoni Jaume-i-Capó,Jose Maria Buades Rubio

【3】Mining Citywide Dengue Spread Patterns in Singapore Through Hotspot Dynamics from Open Web Data
标题：通过开放网络数据的热点动态挖掘新加坡全市范围内的登革热传播模式
链接：https://arxiv.org/abs/2601.12856

作者：Liping Huang,Gaoxi Xiao,Stefan Ma,Hechang Chen,Shisong Tang,Flora Salim
备注：9 pages, 9 figures. It's accepted by WWW 2026 Web4Good Track. To make accessible earlier, authors would like to put it on arxiv before the conference

【4】Exploiting Test-Time Augmentation in Federated Learning for Brain Tumor MRI Classification
标题：利用联邦学习中的测试时间增强进行脑肿瘤MRI分类
链接：https://arxiv.org/abs/2601.12671

作者：Thamara Leandra de Deus Melo,Rodrigo Moreira,Larissa Ferreira Rodrigues Moreira,André Ricardo Backes
备注：21st International Conference on Computer Vision Theory and Applications (VISAPP 2026), 9-11 March 2026, Marbella, Spain

【5】Generalizable Hyperparameter Optimization for Federated Learning on Non-IID Cancer Images
标题：非IID癌症图像联邦学习的可推广超参数优化
链接：https://arxiv.org/abs/2601.12664

作者：Elisa Gonçalves Ribeiro,Rodrigo Moreira,Larissa Ferreira Rodrigues Moreira,André Ricardo Backes
备注：21st International Conference on Computer Vision Theory and Applications (VISAPP 2026), 9-11 March 2026, Marbella, Spain

【6】Explainable Machine Learning for Pediatric Dental Risk Stratification Using Socio-Demographic Determinants
标题：使用社会人口决定因素进行儿科牙科风险分层的可解释机器学习
链接：https://arxiv.org/abs/2601.12405

作者：Manasi Kanade,Abhi Thakkar,Gabriela Fernandes

【7】Automating Parameter Selection in Deep Image Prior for Fluorescence Microscopy Image Denoising via Similarity-Based Parameter Transfer
标题：基于相似性的参数传递的荧光显微图像去噪深度先验参数自动选择
链接：https://arxiv.org/abs/2601.12055

作者：Lina Meyer,Felix Wissel,Tobias Knopp,Susanne Pfefferle,Ralf Fliegert,Maximilian Sandmann,Liana Uebler,Franziska Möckl,Björn-Philipp Diercks,David Lohr,René Werner

【8】Early Linguistic Pattern of Anxiety from Social Media Using Interpretable Linguistic Features: A Multi-Faceted Validation Study with Author-Disjoint Evaluation
标题：使用可解释语言特征的社交媒体焦虑的早期语言模式：一项具有作者不相交评估的多方面验证研究
链接：https://arxiv.org/abs/2601.11758

作者：Arnab Das Utsa
备注：9 figures, more than 1o pages

【9】PSSF: Early osteoarthritis detection using physical synthetic knee X-ray scans and AI radiomics models
标题：PSSF：使用物理合成膝关节X射线扫描和AI放射组学模型进行早期骨关节炎检测
链接：https://arxiv.org/abs/2601.11642

作者：Abbas Alzubaidi,Ali Al-Bayaty
备注：16 pages, 6 figures

【10】onepot CORE -- an enumerated chemical space to streamline drug discovery, enabled by automated small molecule synthesis and AI
标题：onepot CORE --一个通过自动化小分子合成和人工智能实现的简化药物发现的列举化学空间
链接：https://arxiv.org/abs/2601.12603

作者：Andrei S. Tyrin,Brandon Wang,Manuel Muñoz,Samuel H. Foxman,Daniil A. Boiko

【11】Karhunen-Loève Expansion-Based Residual Anomaly Map for Resource-Efficient Glioma MRI Segmentation
标题：基于Karhunen-Loève展开的残余异常图用于资源高效的神经胶质瘤MRI分割
链接：https://arxiv.org/abs/2601.11833

作者：Anthony Hur

蒸馏|知识提取(3篇)

【1】PDFInspect: A Unified Feature Extraction Framework for Malicious Document Detection
标题：PDFInspect：用于恶意文档检测的统一特征提取框架
链接：https://arxiv.org/abs/2601.12866

作者：Sharmila S P
备注：6 pages, 2 figures, paper accepted in COMSNETS 2026 conference

【2】Distilling Time Series Foundation Models for Efficient Forecasting
标题：提炼时间序列基础模型以实现高效预测
链接：https://arxiv.org/abs/2601.12785

作者：Yuqi Li,Kuiye Ding,Chuanguang Yang,Szu-Yu Chen,Yingli Tian
备注：Accepted by ICASSP-2026

【3】jBOT: Semantic Jet Representation Clustering Emerges from Self-Distillation
标题：jBOT：自蒸馏产生的语义喷射表示集群
链接：https://arxiv.org/abs/2601.11719

作者：Ho Fung Tsoi,Dylan Rankin
备注：Under review

聚类(3篇)

【1】Approximation Algorithm for Constrained $k$-Center Clustering: A Local Search Approach
标题：受约束$k$-中心聚集的逼近算法：局部搜索方法
链接：https://arxiv.org/abs/2601.11883

作者：Chaoqi Jia,Longkun Guo,Kewen Liao,Zhigang Lu,Chao Chen,Jason Xue
备注：AAAI-26

【2】Concatenated Matrix SVD: Compression Bounds, Incremental Approximation, and Error-Constrained Clustering
标题：级联矩阵奇异值分解：压缩边界、增量逼近和误差约束集群
链接：https://arxiv.org/abs/2601.11626

作者：Maksym Shamrai

【3】Mixture-of-Experts as Soft Clustering: A Dual Jacobian-PCA Spectral Geometry Perspective
标题：作为软集群的专家混合：二元Jacobian-PCA谱几何观点
链接：https://arxiv.org/abs/2601.11616

作者：Feilong Liu

超分辨率|去噪|去模糊|去雾(1篇)

【1】Physics-Constrained Denoising Autoencoders for Data-Scarce Wildfire UAV Sensing
标题：用于数据稀缺野火无人机传感的物理约束去噪自动编码器
链接：https://arxiv.org/abs/2601.11794

作者：Abdelrahman Ramadan,Zahra Dorbeigi Namaghi,Emily Taylor,Lucas Edwards,Xan Giuliani,David S. McLagan,Sidney Givigi,Melissa Greeff

自动驾驶|车辆|车道检测等(2篇)

【1】Rig-Aware 3D Reconstruction of Vehicle Undercarriages using Gaussian Splatting
标题：利用高斯飞溅进行车辆底盘的Rig-Aware 3D重建
链接：https://arxiv.org/abs/2601.14208

作者：Nitin Kulkarni,Akhil Devarashetti,Charlie Cluss,Livio Forte,Dan Buckmaster,Philip Schneider,Chunming Qiao,Alina Vereshchaka
备注：8 pages, 9 figures, Conference: IEEE International Conference on Machine Learning and Applications 2025 (ICMLA 2025): https://www.icmla-conference.org/icmla25/

【2】Credible CO2 Comparisons: A Machine Learning Approach to Vehicle Powertrain Assessment
标题：可信的二氧化碳比较：车辆动力总成评估的机器学习方法
链接：https://arxiv.org/abs/2601.14022

作者：Rodrigo Pereira David,Luciano Araujo Dourado Filho,Daniel Marques da Silva,João Alfredo Cal-Braz

点云|SLAM|雷达|激光|深度RGBD相关(2篇)

【1】Machine learning model for predicting surface wettability in laser-textured metal alloys
标题：预测激光纹理金属合金表面湿润性的机器学习模型
链接：https://arxiv.org/abs/2601.11661

作者：Mohammad Mohammadzadeh Sanandaji,Danial Ebrahimzadeh,Mohammad Ikram Haider,Yaser Mike Banad,Aleksandar Poleksic,Hongtao Ding
备注：This manuscript has 9 figures and contains 16 pages two column. submitted to journal of laser applications. Under review

【2】Impact of Circuit Depth versus Qubit Count on Variational Quantum Classifiers for Higgs Boson Signal Detection
标题：电路深度与量子比特数对用于Higgs玻色子信号检测的变分量子分类器的影响
链接：https://arxiv.org/abs/2601.11937

作者：Fatih Maulana
备注：13 Pages, 5 Figures, Code and Data Available at: https://github.com/Fatihmaull/higgsboson-detection

联邦学习|隐私保护|加密(5篇)

【1】Federated Balanced Learning
标题：联邦平衡学习
链接：https://arxiv.org/abs/2601.14042

作者：Jiaze Li,Haoran Xu,Wanyi Wu,Changwei Wang,Shuaiguang Li,Jianzhong Ju,Zhenbo Luo,Jian Luan,Youyang Qu,Longxiang Gao,Xudong Yang,Lumin Xing

【2】Fisher-Informed Parameterwise Aggregation for Federated Learning with Heterogeneous Data
标题：基于费舍尔的知情参数化聚合，用于具有异类数据的联邦学习
链接：https://arxiv.org/abs/2601.13608

作者：Zhipeng Chang,Ting He,Wenrui Hao

【3】Federated Learning Under Temporal Drift -- Mitigating Catastrophic Forgetting via Experience Replay
标题：时间漂移下的联邦学习--通过经验回放减轻灾难性遗忘
链接：https://arxiv.org/abs/2601.13456

作者：Sahasra Kokkula,Daniel David,Aaditya Baruah
备注：8 pages, 5 figures. Course project for Neural Networks & Deep Learning COMSW4776 course at Columbia University

【4】Federated Joint Learning for Domain and Class Generalization
标题：用于领域和类概括的联邦联合学习
链接：https://arxiv.org/abs/2601.12253

作者：Haoran Xu,Jiaze Li,Jianzhong Ju,Zhenbo Luo
备注：ICASSP 2026

【5】Federated Learning for the Design of Parametric Insurance Indices under Heterogeneous Renewable Production Losses
标题：非均匀可再生能源生产损失下参数保险指数设计的联邦学习
链接：https://arxiv.org/abs/2601.12178

作者：Fallou Niakh

推理|分析|理解|解释(19篇)

【1】Probabilistic Deep Discriminant Analysis for Wind Blade Segmentation
标题：风叶分割的概率深度鉴别分析
链接：https://arxiv.org/abs/2601.13852

作者：Raül Pérez-Gonzalo,Andreas Espersen,Antonio Agudo
备注：Accepted to ICASSP 2026

【2】TimeART: Towards Agentic Time Series Reasoning via Tool-Augmentation
标题：TimeART：通过工具增强实现抽象时间序列推理
链接：https://arxiv.org/abs/2601.13653

作者：Xingjian Wu,Junkai Lu,Zhengyu Li,Xiangfei Qiu,Jilin Hu,Chenjuan Guo,Christian S. Jensen,Bin Yang

【3】GeoDynamics: A Geometric State-Space Neural Network for Understanding Brain Dynamics on Riemannian Manifolds
标题：GeDynamics：一种用于理解Riemann动量上大脑动力学的几何状态空间神经网络
链接：https://arxiv.org/abs/2601.13570

作者：Tingting Dan,Jiaqi Ding,Guorong Wu
备注：Accepted to NeurIPS 2025

【4】Reasoning is a Modality
标题：推理是一种形态
链接：https://arxiv.org/abs/2601.13562

作者：Zhiguang Liu,Yi Shang
备注：Code access: https://github.com/lz7fd/Reasoning_is_a_Modality

【5】Analysis of Long Range Dependency Understanding in State Space Models
标题：状态空间模型中的长期依赖性理解分析
链接：https://arxiv.org/abs/2601.13048

作者：Srividya Ravikumar,Abhinav Anand,Shweta Verma,Mira Mezini

【6】An efficient heuristic for geometric analysis of cell deformations
标题：单元变形几何分析的有效启发式
链接：https://arxiv.org/abs/2601.12928

作者：Yaima Paz Soto,Silena Herold Garcia,Ximo Gual-Arnau,Antoni Jaume-i-Capó,Manuel González-Hidalgo

【7】SCULPT: Constraint-Guided Pruned MCTS that Carves Efficient Paths for Mathematical Reasoning
标题：SCULPT：约束引导修剪MCTS，为数学推理开辟有效路径
链接：https://arxiv.org/abs/2601.12842

作者：Qitong Fang,Haotian Li,Xu Wang
备注：11 pages, 3 figures. Equal contribution: Qitong Fang and Haotian Li. Corresponding authors: Qitong Fang (fangqitong@student.jlju.edu.cn), Haotian Li (lihaotian@student.jlju.edu.cn), Xu Wang (wangxu@jlju.edu.cn)

【8】Knowledge-Integrated Representation Learning for Crypto Anomaly Detection under Extreme Label Scarcity; Relational Domain-Logic Integration with Retrieval-Grounded Context and Path-Level Explanations
标题：标签极度稀缺下用于加密异常检测的知识集成表示学习;具有检索扎根上下文和路径级解释的关系域逻辑集成
链接：https://arxiv.org/abs/2601.12839

作者：Gyuyeon Na,Minjung Park,Soyoun Kim,Jungbin Shin,Sangmi Chai
备注：Gyuyeon Na, Minjung Park, Soyoun Kim contributed equally to this work

【9】SoundPlot: An Open-Source Framework for Birdsong Acoustic Analysis and Neural Synthesis with Interactive 3D Visualization
标题：SoundPlot：一个用于鸟鸣声学分析和神经合成的开源框架，具有交互式3D可视化
链接：https://arxiv.org/abs/2601.12752

作者：Naqcho Ali Mehdi,Mohammad Adeel,Aizaz Ali Larik

【10】Logic-Guided Multistage Inference for Explainable Multidefendant Judgment Prediction
标题：可解释多被告判决预测的逻辑引导多阶段推理
链接：https://arxiv.org/abs/2601.12688

作者：Xu Zhang,Qinghua Wang,Mengyang Zhao,Fang Wang,Cunquan Qu

【11】Toward Faithful Explanations in Acoustic Anomaly Detection
标题：声学异常检测中的忠实解释
链接：https://arxiv.org/abs/2601.12660

作者：Maab Elrashid,Anthony Deschênes,Cem Subakan,Mirco Ravanelli,Rémi Georges,Michael Morin
备注：Accepted at the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2026. Code: https://github.com/Maab-Nimir/Faithful-Explanations-in-Acoustic-Anomaly-Detection

【12】Streaming Operator Inference for Model Reduction of Large-Scale Dynamical Systems
标题：大规模动态系统模型约简的流运算符推理
链接：https://arxiv.org/abs/2601.12161

作者：Tomoki Koike,Prakash Mohan,Marc T. Henry de Frahan,Julie Bessac,Elizabeth Qian

【13】Process In-Context Learning: Enhancing Mathematical Reasoning via Dynamic Demonstration Insertion
标题：过程上下文学习：通过动态演示插入增强数学推理
链接：https://arxiv.org/abs/2601.11979

作者：Ang Gao,Changshuo Zhang,Xiao Zhang,Deyang Li,Minjun Zhao,Fangchao Liu,Xinyu Zhang

【14】Towards Airborne Object Detection: A Deep Learning Analysis
标题：迈向空中物体检测：深度学习分析
链接：https://arxiv.org/abs/2601.11907

作者：Prosenjit Chatterjee,ANK Zaman

【15】Suspicious Alignment of SGD: A Fine-Grained Step Size Condition Analysis
标题：SGD可疑对齐：细粒度步进条件分析
链接：https://arxiv.org/abs/2601.11789

作者：Shenyang Deng,Boyao Liao,Zhuoli Ouyang,Tianyu Pang,Minhak Song,Yaoqing Yang
备注：The 37th International Conference on Algorithmic Learning Theory

【16】Reasoning Stabilization Point: A Training-Time Signal for Stable Evidence and Shortcut Reliance
标题：推理稳定点：稳定证据和证明可靠性的训练时信号
链接：https://arxiv.org/abs/2601.11625

作者：Sahil Rajesh Dhayalkar
备注：8 pages, Submitted to ACL Rolling Review and is under review

【17】Persistent Sheaf Laplacian Analysis of Protein Stability and Solubility Changes upon Mutation
标题：突变后蛋白质稳定性和溶解度变化的持久性拉普拉斯分析
链接：https://arxiv.org/abs/2601.12219

作者：Yiming Ren,Junjie Wee,Xi Chen,Grace Qian,Guo-Wei Wei

【18】A Kernel Approach for Semi-implicit Variational Inference
标题：半隐式变分推理的核方法
链接：https://arxiv.org/abs/2601.12023

作者：Longlin Yu,Ziheng Cheng,Shiyue Zhang,Cheng Zhang
备注：40 pages, 15 figures. arXiv admin note: substantial text overlap with arXiv:2405.18997

【19】Explainable histomorphology-based survival prediction of glioblastoma, IDH-wildtype
标题：基于组织形态学的可解释的胶质母细胞瘤生存预测，IDH-野生型
链接：https://arxiv.org/abs/2601.11691

作者：Jan-Philipp Redlich,Friedrich Feuerhake,Stefan Nikolin,Nadine Sarah Schaadt,Sarah Teuber-Hanselmann,Joachim Weis,Sabine Luttmann,Andrea Eberle,Christoph Buck,Timm Intemann,Pascal Birnstill,Klaus Kraywinkel,Jonas Ort,Peter Boor,André Homeyer

检测相关(4篇)

【1】Towards Token-Level Text Anomaly Detection
标题：迈向代币级文本异常检测
链接：https://arxiv.org/abs/2601.13644

作者：Yang Cao,Bicheng Yu,Sikun Yang,Ming Liu,Yujiu Yang
备注：WWW 2026

【2】TwoHead-SwinFPN: A Unified DL Architecture for Synthetic Manipulation, Detection and Localization in Identity Documents
标题：TwoHead-SwinFPN：一种用于身份文档中综合操纵、检测和定位的统一DL架构
链接：https://arxiv.org/abs/2601.12895

作者：Chan Naseeb,Adeel Ashraf Cheema,Hassan Sami,Tayyab Afzal,Muhammad Omair,Usman Habib
备注：8 pages

【3】SDCoNet: Saliency-Driven Multi-Task Collaborative Network for Remote Sensing Object Detection
标题：SDCoNet：显著性驱动的遥感目标检测多任务协作网络
链接：https://arxiv.org/abs/2601.12507

作者：Ruo Qi,Linhui Dai,Yusong Qin,Chaolei Yang,Yanshan Li

【4】Machine Learning-Based Framework for Real Time Detection and Early Prediction of Control Valve Stiction in Industrial Control Systems
标题：基于机器学习的工业控制系统中控制阀止动实时检测和早期预测框架
链接：https://arxiv.org/abs/2601.12362

作者：Natthapong Promsricha,Chotirawee Chatpattanasiri,Nuttavut Kerdgongsup,Stavroula Balabani

分类|识别(10篇)

【1】Variational Dual-path Attention Network for CSI-Based Gesture Recognition
标题：基于CSI的变分双路径注意力网络手势识别
链接：https://arxiv.org/abs/2601.13745

作者：N. Zhang
备注：8 pages, 7 figures, 2 tables

【2】Classifiers in High Dimensional Hilbert Metrics
标题：多维Hilbert Buttons中的分类器
链接：https://arxiv.org/abs/2601.13410

作者：Aditya Acharya,Auguste H. Gezalyan,David M. Mount

【3】SASA: Semantic-Aware Contrastive Learning Framework with Separated Attention for Triple Classification
标题：SASA：三重分类注意力分离的语义感知对比学习框架
链接：https://arxiv.org/abs/2601.13035

作者：Xu Xiaodan,Hu Xiaolin
备注：in progress

【4】Dynamic Hand Gesture Recognition for Robot Manipulator Tasks
标题：机器人机械手任务的动态手势识别
链接：https://arxiv.org/abs/2601.12918

作者：Dharmendra Sharma,Peeyush Thakur,Sandeep Gupta,Narendra Kumar Dhar,Laxmidhar Behera

【5】SSVD-O: Parameter-Efficient Fine-Tuning with Structured SVD for Speech Recognition
标题：SSVD-O：采用结构化MVD进行参数高效微调，用于语音识别
链接：https://arxiv.org/abs/2601.12600

作者：Pu Wang,Shinji Watanabe,Hugo Van hamme
备注：Accepted by IEEE ICASSP 2026

【6】Failure Modes in Multi-Hop QA: The Weakest Link Law and the Recognition Bottleneck
标题：多跳QA中的失败模式：最弱的链接定律和识别瓶颈
链接：https://arxiv.org/abs/2601.12499

作者：Meiru Zhang,Zaiqiao Meng,Nigel Collier
备注：preprint

【7】MongoDB Injection Query Classification Model using MongoDB Log files as Training Data
标题：使用MongoDB日志文件作为训练数据的MongoDB注入查询分类模型
链接：https://arxiv.org/abs/2601.11996

作者：Shaunak Perni,Minal Shirodkar,Ramdas Karmalli
备注：24 Pages, 5 Tables, 6 Figures, Journal

【8】Effects of Gabor Filters on Classification Performance of CNNs Trained on a Limited Number of Conditions
标题：Gabor过滤器对有限条件下训练的CNN分类性能的影响
链接：https://arxiv.org/abs/2601.11918

作者：Akito Morita,Hirotsugu Okuno
备注：5 pages, 4 figures, 4 tables

【9】Integrating Temporal Context into Streaming Data for Human Activity Recognition in Smart Home
标题：智能家居中将时间上下文整合到流数据中的人类活动识别
链接：https://arxiv.org/abs/2601.11611

作者：Marina Vicini,Martin Rudorfer,Zhuangzhuang Dai,Luis J. Manso
备注：Accepted to International Conference on Ubiquitous Computing and Ambient Intelligence (UCAmI) 2024

【10】Purification Before Fusion: Toward Mask-Free Speech Enhancement for Robust Audio-Visual Speech Recognition
标题：融合前的净化：迈向无屏蔽语音增强以实现稳健的视听语音识别
链接：https://arxiv.org/abs/2601.12436

作者：Linzhi Wu,Xingyu Zhang,Hao Yuan,Yakun Zhang,Changyan Zheng,Liang Xie,Tiejun Liu,Erwei Yin
备注：Accepted by ICASSP2026

表征(5篇)

【1】Beyond Mapping : Domain-Invariant Representations via Spectral Embedding of Optimal Transport Plans
标题：超越映射：通过最优交通计划的谱嵌入的域不变表示
链接：https://arxiv.org/abs/2601.13350

作者：Abdel Djalil Sad Saoud,Fred Maurice Ngolè Mboula,Hanane Slimani
备注：5 pages, 2 figures

【2】Architecture-Optimization Co-Design for Physics-Informed Neural Networks Via Attentive Representations and Conflict-Resolved Gradients
标题：通过专注表示和预算分解的子索物理信息神经网络的架构优化协同设计
链接：https://arxiv.org/abs/2601.12971

作者：Pancheng Niu,Jun Guo,Qiaolin He,Yongming Chen,Yanchao Shi

【3】ParaMETA: Towards Learning Disentangled Paralinguistic Speaking Styles Representations from Speech
标题：ParaMETA：学习从言语中分离出副语言说话风格的表达
链接：https://arxiv.org/abs/2601.12289

作者：Haowei Lou,Hye-young Paik,Wen Hu,Lina Yao
备注：9 pages, 7 figures, Accepted to AAAI-26 (Main Technical Track)

【4】Learning Longitudinal Health Representations from EHR and Wearable Data
标题：从EHR和可穿戴数据学习纵向健康表示
链接：https://arxiv.org/abs/2601.12227

作者：Yuanyun Zhang,Han Zhou,Li Feng,Yilin Hong,Shi Li

【5】Accelerated MR Elastography Using Learned Neural Network Representation
标题：基于学习神经网络的加速MR弹性成像
链接：https://arxiv.org/abs/2601.11878

作者：Xi Peng

优化|敛散性(15篇)

【1】Optimal L2 Regularization in High-dimensional Continual Linear Regression
标题：高维连续线性回归的最优L2正则化
链接：https://arxiv.org/abs/2601.13844

作者：Gilad Karpel,Edward Moroshko,Ran Levinstein,Ron Meir,Daniel Soudry,Itay Evron
备注：Accepted to ALT 2026

【2】Performance and Complexity Trade-off Optimization of Speech Models During Training
标题：训练期间语音模型的性能和复杂性权衡优化
链接：https://arxiv.org/abs/2601.13704

作者：Esteban Gómez,Tom Bäckström

【3】Self-Improvement as Coherence Optimization: A Theoretical Account
标题：自我完善作为一致性优化：理论解释
链接：https://arxiv.org/abs/2601.13566

作者：Tianyi Qiu,Ahmed Hani Ismail,Zhonghao He,Shi Feng
备注：39 pages

【4】Fairness-informed Pareto Optimization : An Efficient Bilevel Framework
标题：基于公平的帕累托优化：一个有效的二层框架
链接：https://arxiv.org/abs/2601.13448

作者：Sofiane Tanji,Samuel Vaiter,Yassine Laguel

【5】CausationEntropy: Pythonic Optimal Causation Entropy
标题：CausationEntropy：Python最优CausationEntropy
链接：https://arxiv.org/abs/2601.13365

作者：Kevin Slote,Jeremie Fish,Erik Bollt

【6】Distribution-Centric Policy Optimization Dominates Exploration-Exploitation Trade-off
标题：以分销为中心的政策优化主导勘探-开发权衡
链接：https://arxiv.org/abs/2601.12730

作者：Zhaochun Li,Chen Wang,Jionghao Bai,Shisheng Cui,Ge Lan,Zhou Zhao,Yue Wang

【7】Beyond Softmax and Entropy: Improving Convergence Guarantees of Policy Gradients by f-SoftArgmax Parameterization with Coupled Regularization
标题：超越Softmax和Entropy：通过f-SoftArgmax参数化和耦合正规化改善政策对象的收敛保证
链接：https://arxiv.org/abs/2601.12604

作者：Safwan Labbi,Daniil Tiapkin,Paul Mangold,Eric Moulines

【8】Orthogonalized Policy Optimization:Decoupling Sampling Geometry from Optimization Geometry in RLHF
标题：随机化策略优化：RLHF中的最优几何与抽样几何解耦
链接：https://arxiv.org/abs/2601.12415

作者：Wang Zixian

【9】Trainability-Oriented Hybrid Quantum Regression via Geometric Preconditioning and Curriculum Optimization
标题：通过几何预处理和课程优化的面向可训练性的混合量子回归
链接：https://arxiv.org/abs/2601.11942

作者：Qingyu Meng,Yangshuai Wang

【10】Global Optimization By Gradient from Hierarchical Score-Matching Spaces
标题：分层分数匹配空间的梯度进行全局优化
链接：https://arxiv.org/abs/2601.11639

作者：Ming Li

【11】Refined Gradient-Based Temperature Optimization for the Replica-Exchange Monte-Carlo Method
标题：复制品交换Monte-Carlo方法的基于精细物质的温度优化
链接：https://arxiv.org/abs/2601.13542

作者：Tatsuya Miyata,Shunta Arai,Satoshi Takabe
备注：15 pages

【12】Small Gradient Norm Regret for Online Convex Optimization
标题：在线凸优化的小梯度范数后悔算法
链接：https://arxiv.org/abs/2601.13519

作者：Wenzhi Gao,Chang He,Madeleine Udell

【13】BiCoLoR: Communication-Efficient Optimization with Bidirectional Compression and Local Training
标题：BiCoLoR：通过双向压缩和本地训练实现通信高效优化
链接：https://arxiv.org/abs/2601.12400

作者：Laurent Condat,Artavazd Maranjyan,Peter Richtárik

【14】On the Provable Suboptimality of Momentum SGD in Nonstationary Stochastic Optimization
标题：关于非平稳随机优化中动量SGD的可证次优性
链接：https://arxiv.org/abs/2601.12238

作者：Sharan Sahu,Cameron J. Hogan,Martin T. Wells
备注：70 pages, 4 figures, 2 tables

【15】Offline Policy Learning with Weight Clipping and Heaviside Composite Optimization
标题：通过减肥和Heaviside复合优化进行离线政策学习
链接：https://arxiv.org/abs/2601.12117

作者：Jingren Liu,Hanzhang Qin,Junyi Liu,Mabel C. Chou,Jong-Shi Pang

预测|估计(19篇)

【1】PAtt: A Pattern Attention Network for ETA Prediction Using Historical Speed Profiles
标题：Patt：使用历史速度曲线预测埃塔的模式注意力网络
链接：https://arxiv.org/abs/2601.13793

作者：ByeoungDo Kim,JunYeop Na,Kyungwook Tak,JunTae Kim,DongHyeon Kim,Duckky Kim
备注：7 pages, 3 figures, ITSC 2025, to be published

【2】vLinear: A Powerful Linear Model for Multivariate Time Series Forecasting
标题：vLinear：用于多元时间序列预测的强大线性模型
链接：https://arxiv.org/abs/2601.13768

作者：Wenzhen Yue,Ruohao Guo,Ji Shi,Zihan Hao,Shiyu Hu,Xianghua Ying

【3】EEG-Titans: Long-Horizon Seizure Forecasting via Dual-Branch Attention and Neural Memory
标题：脑电巨人：通过双分支注意力和神经记忆进行癫痫发作的长期预测
链接：https://arxiv.org/abs/2601.13748

作者：Tien-Dat Pham,Xuan-The Tran

【4】Does Privacy Always Harm Fairness? Data-Dependent Trade-offs via Chernoff Information Neural Estimation
标题：隐私总是损害公平吗？通过叠加信息神经估计的数据相关权衡
链接：https://arxiv.org/abs/2601.13698

作者：Arjun Nichani,Hsiang Hsu,Chun-Fu,Chen,Haewon Jeong

【5】Optimizing Parallel Schemes with Lyapunov Exponents and kNN-LLE Estimation
标题：用李雅普诺夫指数和kNN-LLE估计优化并行格式
链接：https://arxiv.org/abs/2601.13604

作者：Mudassir Shams,Andrei Velichko,Bruno Carpentieri
备注：25 pages, 9 figures, 10 tables

【6】Bridging the Gap Between Estimated and True Regret Towards Reliable Regret Estimation in Deep Learning based Mechanism Design
标题：在基于深度学习的机制设计中弥合估计遗憾和真实遗憾之间的差距，实现可靠的遗憾估计
链接：https://arxiv.org/abs/2601.13489

作者：Shuyuan You,Zhiqiang Zhuang,Kewen Wang,Zhe Wang

【7】TrustEnergy: A Unified Framework for Accurate and Reliable User-level Energy Usage Prediction
标题：TrustEnergy：准确可靠的用户级能源使用预测的统一框架
链接：https://arxiv.org/abs/2601.13422

作者：Dahai Yu,Rongchao Xu,Dingyi Zhuang,Yuheng Bu,Shenhao Wang,Guang Wang

【8】Trend-Adjusted Time Series Models with an Application to Gold Price Forecasting
标题：趋势调整时间序列模型及其在黄金价格预测中的应用
链接：https://arxiv.org/abs/2601.12706

作者：Sina Kazemdehbashi

【9】Topology-Aware Multiscale Mixture of Experts for Efficient Molecular Property Prediction
标题：用于高效分子性质预测的具有布局感知的多尺度专家混合
链接：https://arxiv.org/abs/2601.12637

作者：Long D. Nguyen,Kelin Xia,Binh P. Nguyen

【10】Life, Machine Learning, and the Search for Habitability: Predicting Biosignature Fluxes for the Habitable Worlds Observatory
标题：生命、机器学习和寻找宜居性：预测宜居世界天文台的生物特征通量
链接：https://arxiv.org/abs/2601.12557

作者：Mark Moussa,Amber V. Young,Brianna Isola,Vasuda Trehan,Michael D. Himes,Nicholas Wogan,Giada Arney
备注：8 pages, 4 figures. Submitted and accepted in AAAI-26 (IAAI Emerging Applications track)

【11】IceWatch: Forecasting Glacial Lake Outburst Floods (GLOFs) using Multimodal Deep Learning
标题：IceWatch：使用多模式深度学习预测冰川湖爆发洪水（GIBF）
链接：https://arxiv.org/abs/2601.12330

作者：Zuha Fatima,Muhammad Anser Sohaib,Muhammad Talha,Ayesha Kanwal,Sidra Sultana,Nazia Perwaiz

【12】Distribution Shift Is Key to Learning Invariant Prediction
标题：分布转变是学习不变预测的关键
链接：https://arxiv.org/abs/2601.12296

作者：Hong Zheng,Fei Teng

【13】FutureX-Pro: Extending Future Prediction to High-Value Vertical Domains
标题：FutureX-Pro：将未来预测扩展到高价值垂直领域
链接：https://arxiv.org/abs/2601.12259

作者：Jiashuo Liu,Siyuan Chen,Zaiyuan Wang,Zhiyuan Zeng,Jiacheng Guo,Liang Hu,Lingyue Yin,Suozhi Huang,Wenxin Hao,Yang Yang,Zerui Cheng,Zixin Yao,Lingyue Yin,Haoxin Liu,Jiayi Cheng,Yuzhen Li,Zezhong Ma,Bingjie Wang,Bingsen Qiu,Xiao Liu,Zeyang Zhang,Zijian Liu,Jinpeng Wang,Mingren Yin,Tianci He,Yali Liao,Yixiao Tian,Zhenwei Zhu,Anqi Dai,Ge Zhang,Jingkai Liu,Kaiyuan Zhang,Wenlong Wu,Xiang Gao,Xinjie Chen,Zhixin Yao,Zhoufutu Wen,B. Aditya Prakash,Jose Blanchet,Mengdi Wang,Nian Si,Wenhao Huang
备注：21 pages

【14】Shapelets-Enriched Selective Forecasting using Time Series Foundation Models
标题：使用时间序列基础模型的Shapelets丰富的选择性预测
链接：https://arxiv.org/abs/2601.11821

作者：Shivani Tomar,Seshu Tirupathi,Elizabeth Daly,Ivana Dusparic
备注：Accepted by the AAAI-26 Workshop on Artificial Intelligence for Time Series Analysis (AI4TS)

【15】A Review on Machine Learning Approaches for the Prediction of Glucose Levels and Hypogylcemia
标题：预测血糖水平和低血糖的机器学习方法综述
链接：https://arxiv.org/abs/2601.11615

作者：Beyza Cinar,Louisa van den Boom,Maria Maleshkova

【16】Auxiliary-predicted Compress Memory Model(ApCM Model): A Neural Memory Storage Model Based on Invertible Compression and Learnable Prediction
标题：辅助预测压缩记忆模型（ApCM模型）：基于可逆压缩和可学习预测的神经记忆存储模型
链接：https://arxiv.org/abs/2601.11609

作者：Weinuo Ou
备注：9 pages, 7 figures

【17】Intermittent time series forecasting: local vs global models
标题：间歇性时间序列预测：本地与全球模型
链接：https://arxiv.org/abs/2601.14031

作者：Stefano Damato,Nicolò Rubattu,Dario Azzimonti,Giorgio Corani
备注：Submitted to Data Mining and Knowledge Discovery

【18】Unified Unbiased Variance Estimation for MMD: Robust Finite-Sample Performance with Imbalanced Data and Exact Acceleration under Null and Alternative Hypotheses
标题：MMD的统一无偏方差估计：数据不平衡和可替代假设下的精确加速的稳健伪样本性能
链接：https://arxiv.org/abs/2601.13874

作者：Shijie Zhong,Jiangfeng Fu,Yikun Yang

【19】Approximate full conformal prediction in RKHS
标题：RKHS中的近似全适形预测
链接：https://arxiv.org/abs/2601.13102

作者：Davidson Lova Razafindrakoto,Alain Celisse,Jérôme Lacaille

其他神经网络|深度学习|模型|建模(57篇)

【1】Q-learning with Adjoint Matching
标题：伴随匹配的Q学习
链接：https://arxiv.org/abs/2601.14234

作者：Qiyang Li,Sergey Levine
备注：32 pages, 8 figures, 7 tables

【2】ConceptCaps -- a Distilled Concept Dataset for Interpretability in Music Models
标题：ConceptCaps --音乐模型可解释性的提炼概念数据集
链接：https://arxiv.org/abs/2601.14157

作者：Bruno Sienkiewicz,Łukasz Neumann,Mateusz Modrzejewski

【3】Causal feature selection framework for stable soft sensor modeling based on time-delayed cross mapping
标题：基于延时交叉映射的稳定软测量建模因果特征选择框架
链接：https://arxiv.org/abs/2601.14099

作者：Shi-Shun Chen,Xiao-Yang Li,Enrico Zio

【4】'1'-bit Count-based Sorting Unit to Reduce Link Power in DNN Accelerators
标题：“1”位基于计数的排序单元可降低DNN加速器中的链路功率
链接：https://arxiv.org/abs/2601.14087

作者：Ruichi Han,Yizhi Chen,Tong Lei,Jordi Altayo Gonzalez,Ahmed Hemani
备注：Accepted for oral presentation at the 2026 VLSI Symposium on Technology, Systems and Applications (VLSI TSA) on April 13-17, 2026, at the Ambassador Hotel, Hsinchu, Taiwan

【5】A universal linearized subspace refinement framework for neural networks
标题：神经网络的通用线性化子空间细化框架
链接：https://arxiv.org/abs/2601.13989

作者：Wenbo Cao,Weiwei Zhang

【6】Towards Effective Negation Modeling in Joint Audio-Text Models for Music
标题：音乐联合音频文本模型中的有效否定建模
链接：https://arxiv.org/abs/2601.13931

作者：Yannis Vasilakis,Rachel Bittner,Johan Pauwels
备注：Accepted at IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2026

【7】Discriminant Learning-based Colorspace for Blade Segmentation
标题：基于判别学习的颜色空间叶片分割
链接：https://arxiv.org/abs/2601.13816

作者：Raül Pérez-Gonzalo,Andreas Espersen,Antonio Agudo
备注：Accepted to ICASSP 2026

【8】Autoregressive deep learning for real-time simulation of soft tissue dynamics during virtual neurosurgery
标题：自回归深度学习用于虚拟神经外科手术期间软组织动力学的实时模拟
链接：https://arxiv.org/abs/2601.13676

作者：Fabian Greifeneder,Wolfgang Fenz,Benedikt Alkin,Johannes Brandstetter,Michael Giretzlehner,Philipp Moser

【9】An Elementary Approach to Scheduling in Generative Diffusion Models
标题：生成扩散模型中调度的一种基本方法
链接：https://arxiv.org/abs/2601.13602

作者：Qiang Sun,H. Vincent Poor,Wenyi Zhang

【10】Diffusion In Diffusion: Breaking the Autoregressive Bottleneck in Block Diffusion Models
标题：扩散中的扩散：打破块扩散模型中的自回归瓶颈
链接：https://arxiv.org/abs/2601.13599

作者：Linrui Ma,Yufei Cui,Kai Han,Yunhe Wang
备注：Work In Progress

【11】Machine learning based radiative parameterization scheme and its performance in operational reforecast experiments
标题：基于机器学习的辐射参数化方案及其业务再预报试验
链接：https://arxiv.org/abs/2601.13592

作者：Hao Jing,Sa Xiao,Haoyu Li,Huadong Xiao,Wei Xue

【12】Behavior Knowledge Merge in Reinforced Agentic Models
标题：强化统计模型中的行为知识融合
链接：https://arxiv.org/abs/2601.13572

作者：Xiangchi Yuan,Dachuan Shi,Chunhui Zhang,Zheyuan Liu,Shenglong Yao,Soroush Vosoughi,Wenke Lee

【13】Quantum Qualifiers for Neural Network Model Selection in Hadronic Physics
标题：强子物理中神经网络模型选择的量子简化器
链接：https://arxiv.org/abs/2601.13463

作者：Brandon B. Le,D. Keller
备注：12 pages, 5 figures. Proceedings for the 26th International Symposium on Spin Physics (SPIN2025), September 21-26, 2025; Qingdao, Shandong, China

【14】BladeSDF : Unconditional and Conditional Generative Modeling of Representative Blade Geometries Using Signed Distance Functions
标题：BladeSDF：使用符号距离函数对代表性叶片几何图形进行无条件和条件生成建模
链接：https://arxiv.org/abs/2601.13445

作者：Ashish S. Nair,Sandipp Krishnan Ravi,Itzel Salgado,Changjie Sun,Sayan Ghosh,Liping Wang

【15】MOSLD-Bench: Multilingual Open-Set Learning and Discovery Benchmark for Text Categorization
标题：MOSLD-Bench：文本分类的多语言开放集学习和发现基准
链接：https://arxiv.org/abs/2601.13437

作者：Adriana-Valentina Costache,Daria-Nicoleta Dragomir,Silviu-Florin Gheorghe,Eduard Poesina,Paul Irofti,Radu Tudor Ionescu

【16】On the Relation of State Space Models and Hidden Markov Models
标题：状态空间模型与隐马尔科夫模型的关系
链接：https://arxiv.org/abs/2601.13357

作者：Aydin Ghojogh,M. Hadi Sepanj,Benyamin Ghojogh

【17】MultiST: A Cross-Attention-Based Multimodal Model for Spatial Transcriptomic
标题：MultiST：一种基于交叉注意力的空间转录组学多模式模型
链接：https://arxiv.org/abs/2601.13331

作者：Wei Wang,Quoc-Toan Ly,Chong Yu,Jun Bai

【18】Paid Voices vs. Public Feeds: Interpretable Cross-Platform Theme Modeling of Climate Discourse
标题：付费声音与公共提要：气候话语的可解释跨平台主题建模
链接：https://arxiv.org/abs/2601.13317

作者：Samantha Sudhoff,Pranav Perumal,Zhaoqing Wu,Tunazzina Islam

【19】Verifying Local Robustness of Pruned Safety-Critical Networks
标题：删除安全关键网络的本地鲁棒性
链接：https://arxiv.org/abs/2601.13303

作者：Minh Le,Phuong Cao

【20】Deep Neural networks for solving high-dimensional parabolic partial differential equations
标题：用于求解多维方程的深度神经网络
链接：https://arxiv.org/abs/2601.13256

作者：Wenzhong Zhang,Zhenyuan Hu,Wei Cai,George EM Karniadakis

【21】Aligning Agentic World Models via Knowledgeable Experience Learning
标题：通过丰富的经验学习调整宏观世界模型
链接：https://arxiv.org/abs/2601.13247

作者：Baochang Ren,Yunzhi Yao,Rui Sun,Shuofei Qiao,Ningyu Zhang,Huajun Chen
备注：Ongoing work

【22】Do Instruction-Tuned Models Always Perform Better Than Base Models? Evidence from Math and Domain-Shifted Benchmarks
标题：指令调整模型总是比基本模型表现得更好吗？来自数学和领域转移基准的证据
链接：https://arxiv.org/abs/2601.13244

作者：Prateek Munjal,Clement Christophe,Ronnie Rajan,Praveenkumar Kanithi

【23】Training instability in deep learning follows low-dimensional dynamical principles
标题：深度学习中的训练不稳定性遵循低维动态原理
链接：https://arxiv.org/abs/2601.13160

作者：Zhipeng Zhang,Zhenjie Yao,Kai Li,Lei Yang

【24】RM -RF: Reward Model for Run-Free Unit Test Evaluation
标题：RM -RF：免运行单元测试评估的奖励模型
链接：https://arxiv.org/abs/2601.13097

作者：Elena Bruches,Daniil Grebenkin,Mikhail Klementev,Vadim Alperovich,Roman Derunets,Dari Baturova,Georgy Mkrtchyan,Oleg Sedukhin,Ivan Bondarenko,Nikolay Bushkov,Stanislav Moiseev
备注：This paper has been accepted for publication at the 33rd IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER 2026)

【25】PASs-MoE: Mitigating Misaligned Co-drift among Router and Experts via Pathway Activation Subspaces for Continual Learning
标题：PAs-MoE：通过路径激活子空间缓解路由器和专家之间的不对齐协同漂移以实现持续学习
链接：https://arxiv.org/abs/2601.13020

作者：Zhiyan Hou,Haiyun Guo,Haokai Ma,Yandu Sun,Yonghui Yang,Jinqiao Wang

【26】Deterministic Dynamics of Sampling Processes in Score-Based Diffusion Models with Multiplicative Noise Conditioning
标题：具有乘性噪音条件的基于分数的扩散模型中抽样过程的确定性动力学
链接：https://arxiv.org/abs/2601.12965

作者：Doheon Kim

【27】Online Continual Learning for Time Series: a Natural Score-driven Approach
标题：时间序列在线持续学习：自然分数驱动的方法
链接：https://arxiv.org/abs/2601.12931

作者：Edoardo Urettini,Daniele Atzeni,Ioanna-Yvonni Tsaknaki,Antonio Carta

【28】Fisher-Orthogonal Projected Natural Gradient Descent for Continual Learning
标题：用于连续学习的费舍尔-垂直投影自然梯度下降
链接：https://arxiv.org/abs/2601.12816

作者：Ishir Garg,Neel Kolhe,Andy Peng,Rohan Gopalam

【29】SL-CBM: Enhancing Concept Bottleneck Models with Semantic Locality for Better Interpretability
标题：SL-CBM：利用语义局部性增强概念瓶颈模型以提高解释性
链接：https://arxiv.org/abs/2601.12804

作者：Hanwei Zhang,Luo Cheng,Rui Wen,Yang Zhang,Lijun Zhang,Holger Hermanns

【30】P2L-CA: An Effective Parameter Tuning Framework for Rehearsal-Free Multi-Label Class-Incremental Learning
标题：P2 L-CA：免排练多标签类增量学习的有效参数调整框架
链接：https://arxiv.org/abs/2601.12714

作者：Songlin Dong,Jiangyang Li,Chenhao Ding,Zhiheng Ma,Haoyu Luo,Yuhang He,Yihong Gong
备注：12 pages, 5 figures

【31】Adaptively trained Physics-informed Radial Basis Function Neural Networks for Solving Multi-asset Option Pricing Problems
标题：自适应训练的物理信息放射性基函数神经网络用于解决多资产期权定价问题
链接：https://arxiv.org/abs/2601.12704

作者：Yan Ma,Yumeng Ren
备注：30 pages,16 figures

【32】Learning Deterministic Finite-State Machines from the Prefixes of a Single String is NP-Complete
标题：从单个字符串的后缀学习确定性伪状态机是NP完全的
链接：https://arxiv.org/abs/2601.12621

作者：Radu Cosmin Dumitru,Ryo Yoshinaka,Ayumi Shinohara
备注：12 pages, 4 figures

【33】Dissecting Linear Recurrent Models: How Different Gating Strategies Drive Selectivity and Generalization
标题：剖析线性回归模型：不同的门控策略如何推动选择性和概括性
链接：https://arxiv.org/abs/2601.12598

作者：Younes Bouhadjar,Maxime Fabre,Felix Schmidt,Emre Neftci
备注：11 pages, 4 figures and 4 tables

【34】Learning Relativistic Geodesics and Chaotic Dynamics via Stabilized Lagrangian Neural Networks
标题：通过稳定拉格朗日神经网络学习相对论测地学和混乱动力学
链接：https://arxiv.org/abs/2601.12519

作者：Abdullah Umut Hamzaogullari,Arkadas Ozakin
备注：21 pages

【35】Semidefinite Programming for Quantum Channel Learning
标题：量子信道学习的半定规划方法
链接：https://arxiv.org/abs/2601.12502

作者：Mikhail Gennadievich Belov,Victor Victorovich Dubov,Vadim Konstantinovich Ivanov,Alexander Yurievich Maslov,Olga Vladimirovna Proshina,Vladislav Gennadievich Malyshkin

【36】Statistical-Neural Interaction Networks for Interpretable Mixed-Type Data Imputation
标题：用于可解释混合型数据插补的统计-神经交互网络
链接：https://arxiv.org/abs/2601.12380

作者：Ou Deng,Shoji Nishimura,Atsushi Ogihara,Qun Jin

【37】Ordered Local Momentum for Asynchronous Distributed Learning under Arbitrary Delays
标题：任意延迟下同步分布式学习的有序局部动量
链接：https://arxiv.org/abs/2601.12322

作者：Chang-Wei Shi,Shi-Shang Wang,Wu-Jun Li

【38】Machine Learning as a Service (MLaaS) Dataset Generator Framework for IoT Environments
标题：适用于物联网环境的机器学习即服务（MLaSaaS）数据集生成器框架
链接：https://arxiv.org/abs/2601.12305

作者：Deepak Kanneganti,Sajib Mistry,Sheik Fattah,Joshua Boland,Aneesh Krishna

【39】Wavelet-Driven Masked Multiscale Reconstruction for PPG Foundation Models
标题：JPEG基础模型的子波驱动掩蔽多尺度重建
链接：https://arxiv.org/abs/2601.12215

作者：Megha Thukral,Cyrus Tanade,Simon A. Lee,Juhyeon Lee,Hao Zhou,Keum San Chun,Migyeong Gwak,Viswam Nathan,Md Mahbubur Rahman,Li Zhu,Mehrab Bin Morshed,Subramaniam Venkatraman,Sharanya Arcot Desai

【40】Threshold Differential Attention for Sink-Free, Ultra-Sparse, and Non-Dispersive Language Modeling
标题：无水槽、超稀疏和非分散语言建模的阈值差异注意力
链接：https://arxiv.org/abs/2601.12145

作者：Xingyue Huang,Xueying Ding,Mingxuan Ju,Yozen Liu,Neil Shah,Tong Zhao

【41】Learning to Factorize and Adapt: A Versatile Approach Toward Universal Spatio-Temporal Foundation Models
标题：学习分解和适应：通用时空基础模型的通用方法
链接：https://arxiv.org/abs/2601.12083

作者：Siru Zhong,Junjie Qiu,Yangyu Wu,Yiqiu Liu,Yuanpeng He,Zhongwen Rao,Bin Yang,Chenjuan Guo,Hao Xu,Yuxuan Liang
备注：This is an extended version of the paper presented at NeurIPS 2025. Code available at https://github.com/CityMind-Lab/FactoST

【42】Kernel-Based Learning of Safety Barriers
标题：基于核心的安全障碍学习
链接：https://arxiv.org/abs/2601.12002

作者：Oliver Schön,Zhengang Zhong,Sadegh Soudjani
备注：44 pages, 9 figures

【43】Industry-Aligned Granular Topic Modeling
标题：面向行业的粒度主题建模
链接：https://arxiv.org/abs/2601.11762

作者：Sae Young Moon,Myeongjun Erik Jang,Haoyan Luo,Chunyang Xiao,Antonios Georgiadis,Fran Silavong

【44】SpaRRTa: A Synthetic Benchmark for Evaluating Spatial Intelligence in Visual Foundation Models
标题：SpaRRTa：评估视觉基础模型中空间智能的综合基准
链接：https://arxiv.org/abs/2601.11729

作者：Turhan Can Kargin,Wojciech Jasiński,Adam Pardyl,Bartosz Zieliński,Marcin Przewięźlikowski
备注：Project page is available at https://sparrta.gmum.net/

【45】Distill-then-Replace: Efficient Task-Specific Hybrid Attention Model Construction
标题：提取然后替换：高效的特定任务混合注意力模型构建
链接：https://arxiv.org/abs/2601.11667

作者：Xiaojie Xia,Huigang Zhang,Chaoliang Zhong,Jun Sun,Yusuke Oishi

【46】Size is Not the Solution: Deformable Convolutions for Effective Physics Aware Deep Learning
标题：大小不是解决方案：可变形卷积实现有效物理感知深度学习
链接：https://arxiv.org/abs/2601.11657

作者：Jack T. Beerman,Shobhan Roy,H. S. Udaykumar,Stephen S. Baek

【47】Verifying Physics-Informed Neural Network Fidelity using Classical Fisher Information from Differentiable Dynamical System
标题：使用来自可微动力系统的经典Fisher信息来验证物理信息的神经网络保真度
链接：https://arxiv.org/abs/2601.11638

作者：Josafat Ribeiro Leal Filho,Antônio Augusto Fröhlich
备注：This paper has been submitted and is currently under review at IEEE Transactions on Neural Networks and Learning Systems (TNNLS)

【48】Enhancing Model Context Protocol (MCP) with Context-Aware Server Collaboration
标题：通过上下文感知服务器协作增强模型上下文协议（HCP）
链接：https://arxiv.org/abs/2601.11595

作者：Meenakshi Amulya Jayanti,X. Y. Han

【49】Deep Learning Approaches to Quantum Error Mitigation
标题：量子错误缓解的深度学习方法
链接：https://arxiv.org/abs/2601.14226

作者：Leonardo Placidi,Ifan Williams,Enrico Rinaldi,Daniel Mills,Cristina Cîrstoiu,Vanya Eccles,Ross Duncan
备注：48 pages

【50】Labels or Preferences? Budget-Constrained Learning with Human Judgments over AI-Generated Outputs
标题：标签还是偏好？对人工智能生成的输出进行人类判断的预算限制学习
链接：https://arxiv.org/abs/2601.13458

作者：Zihan Dong,Ruijia Wu,Linjun Zhang

【51】Improving Geopolitical Forecasts with Bayesian Networks
标题：利用Bayesian网络改进地缘政治预测
链接：https://arxiv.org/abs/2601.13362

作者：Matthew Martin
备注：34 pages, 3 figures

【52】Polychronous Wave Computing: Timing-Native Address Selection in Spiking Networks
标题：多元波计算：尖峰网络中的定时本地地址选择
链接：https://arxiv.org/abs/2601.13079

作者：Natalila G. Berloff
备注：23 pages, Supplementary Materials are available at https://www.damtp.cam.ac.uk/user/ngb23/publications/SM_PWC.pdf

【53】A Theory of Diversity for Random Matrices with Applications to In-Context Learning of Schrödinger Equations
标题：随机矩阵的多样性理论及其在薛定汉方程上下文学习中的应用
链接：https://arxiv.org/abs/2601.12587

作者：Frank Cole,Yulong Lu,Shaurya Sehgal

【54】Bone-conduction Guided Multimodal Speech Enhancement with Conditional Diffusion Models
标题：基于条件扩散模型的骨导引导多模式语音增强
链接：https://arxiv.org/abs/2601.12354

作者：Sina Khanagha,Bunlong Lay,Timo Gerkmann
备注：Accepted to IEEE ICASSP 2026

【55】A New Strategy for Artificial Intelligence: Training Foundation Models Directly on Human Brain Data
标题：人工智能的新策略：直接根据人脑数据训练基金会模型
链接：https://arxiv.org/abs/2601.12053

作者：Maël Donoso

【56】Quantum Kernel Machine Learning for Autonomous Materials Science
标题：用于自主材料科学的量子核机器学习
链接：https://arxiv.org/abs/2601.11775

作者：Felix Adams,Daiwei Zhu,David W. Steuerman,A. Gilad Kusne,Ichiro Takeuchi

【57】AllShowers: One model for all calorimeter showers
标题：AllShowers：适用于所有热量计淋浴的一种型号
链接：https://arxiv.org/abs/2601.11716

作者：Thorsten Buss,Henry Day-Hall,Frank Gaede,Gregor Kasieczka,Katja Krüger

其他(74篇)

【1】APEX-Agents
标题：APEX-代理
链接：https://arxiv.org/abs/2601.14242

作者：Bertie Vidgen,Austin Mann,Abby Fennelly,John Wright Stanly,Lucas Rothman,Marco Burstein,Julien Benchek,David Ostrofsky,Anirudh Ravichandran,Debnil Sur,Neel Venugopal,Alannah Hsia,Isaac Robinson,Calix Huang,Olivia Varones,Daniyal Khan,Michael Haines,Zach Richards,Chirag Mahapatra,Brendan Foody,Osvald Nitski

【2】Differentiated Pickup Point Offering for Emission Reduction in Last-Mile Delivery
标题：差异化的提货点服务，可减少最后一英里交付的排放
链接：https://arxiv.org/abs/2601.14196

作者：Albina Galiullina,Wouter van Heeswijk,Tom van Woensel

【3】Penalizing Localized Dirichlet Energies in Low Rank Tensor Products
标题：惩罚低阶张量产品中的局部Dirichlet能量
链接：https://arxiv.org/abs/2601.14173

作者：Paris A. Karakasis,Nicholas D. Sidiropoulos
备注：19 pages

【4】Universal Approximation Theorem for Input-Connected Multilayer Perceptrons
标题：输入连通多层感知器的普适逼近定理
链接：https://arxiv.org/abs/2601.14026

作者：Vugar Ismailov
备注：18 pages, 2 figures, 31 references

【5】Auditory Brain Passage Retrieval: Cross-Sensory EEG Training for Neural Information Retrieval
标题：听觉脑通道检索：用于神经信息检索的跨感官脑电训练
链接：https://arxiv.org/abs/2601.14001

作者：Niall McGuire,Yashar Moshfeghi
备注：Accepted At ECIR 2026

【6】Harmonizing the Deep: A Unified Information Pipeline for Robust Marine Biodiversity Assessment Across Heterogeneous Domains
标题：协调深海：跨异类领域稳健海洋生物多样性评估的统一信息管道
链接：https://arxiv.org/abs/2601.13975

作者：Marco Piccolo,Qiwei Han,Astrid van Toor,Joachim Vanneste
备注：9 pages, 4 figures 8 tables

【7】Differentiable Logic Synthesis: Spectral Coefficient Selection via Sinkhorn-Constrained Composition
标题：可微逻辑综合：通过辛霍恩约束合成的谱系数选择
链接：https://arxiv.org/abs/2601.13953

作者：Gorgi Pavlov
备注：35 pages, 22 figures. Code available at https://github.com/gogipav14/spectral-llm

【8】Efficient Coordination with the System-Level Shared State: An Embodied-AI Native Modular Framework
标题：与系统级共享状态的有效协调：一个竞争性AI原生模块化框架
链接：https://arxiv.org/abs/2601.13945

作者：Yixuan Deng,Tongrun Wu,Donghao Wu,Zeyu Wei,Jiayuan Wang,Zhenglong Sun,Yuqing Tang,Xiaoqiang Ji

【9】TrackletGPT: A Language-like GPT Framework for White Matter Tract Segmentation
标题：TrackletGPT：用于白质束分割的类似迷宫的GPT框架
链接：https://arxiv.org/abs/2601.13935

作者：Anoushkrit Goel,Simroop Singh,Ankita Joshi,Ranjeet Ranjan Jha,Chirag Ahuja,Aditya Nigam,Arnav Bhavsar
备注：Accepted at 23rd IEEE International Symposium on Biomedical Imaging (ISBI), 2026

【10】TractRLFusion: A GPT-Based Multi-Critic Policy Fusion Framework for Fiber Tractography
标题：TractRL Fusion：基于GPT的光纤牵引成像多批评政策融合框架
链接：https://arxiv.org/abs/2601.13897

作者：Ankita Joshi,Ashutosh Sharma,Anoushkrit Goel,Ranjeet Ranjan Jha,Chirag Ahuja,Arnav Bhavsar,Aditya Nigam
备注：Accepted at 23rd IEEE International Symposium on Biomedical Imaging (ISBI), 2026

【11】Inverting Self-Organizing Maps: A Unified Activation-Based Framework
标题：倒置自组织地图：统一的基于激活的框架
链接：https://arxiv.org/abs/2601.13851

作者：Alessandro Londei,Matteo Benati,Denise Lanzieri,Vittorio Loreto

【12】Virtual Urbanism: An AI-Driven Framework for Quantifying Urban Identity. A Tokyo-Based Pilot Study Using Diffusion-Generated Synthetic Environments
标题：虚拟城市主义：量化城市身份的人工智能驱动框架。使用扩散生成合成环境的东京试点研究
链接：https://arxiv.org/abs/2601.13846

作者：Glinskaya Maria

【13】Insight: Interpretable Semantic Hierarchies in Vision-Language Encoders
标题：洞察：视觉语言编码器中的可解释语义分层
链接：https://arxiv.org/abs/2601.13798

作者：Kai Wittenmayer,Sukrut Rao,Amin Parchami-Araghi,Bernt Schiele,Jonas Fischer
备注：32 pages, 24 figures, 3 tables

【14】Orthogonium : A Unified, Efficient Library of Orthogonal and 1-Lipschitz Building Blocks
标题：钼酸铵：一个统一、高效的正交和1-Lipschitz构建模块库
链接：https://arxiv.org/abs/2601.13776

作者：Thibaut Boissin,Franck Mamalet,Valentin Lafargue,Mathieu Serrurier

【15】Attention-space Contrastive Guidance for Efficient Hallucination Mitigation in LVLMs
标题：有效缓解LVLM幻觉的注意空间对比指南
链接：https://arxiv.org/abs/2601.13707

作者：Yujin Jo,Sangyoon Bae,Taesup Kim

【16】Quadratic Upper Bound for Boosting Robustness
标题：提高稳健性的二次上界
链接：https://arxiv.org/abs/2601.13645

作者：Euijin You,Hyang-Won Lee
备注：Accepted at ICML 2025. Published in PMLR 267:72656-72676

【17】FG-OrIU: Towards Better Forgetting via Feature-Gradient Orthogonality for Incremental Unlearning
标题：FG-OrIU：通过增量遗忘的学习-梯度异向性实现更好的遗忘
链接：https://arxiv.org/abs/2601.13578

作者：Qian Feng,JiaHang Tu,Mintong Kang,Hanbin Zhao,Chao Zhang,Hui Qian
备注：This paper has been accepted by ICCV 2025. code: \url{https://github.com/RAIAN08/FG-OrIU}

【18】Multi-objective fluorescent molecule design with a data-physics dual-driven generative framework
标题：具有数据物理双重驱动生成框架的多目标荧光分子设计
链接：https://arxiv.org/abs/2601.13564

作者：Yanheng Li,Zhichen Pu,Lijiang Yang,Zehao Zhou,Yi Qin Gao
备注：Total 43 pages: 32 pages Main Text + 11 pages SI

【19】ButterflyMoE: Sub-Linear Ternary Experts via Structured Butterfly Orbits
标题：ButterflyMoE：通过结构化蝴蝶轨道的次线性三元专家
链接：https://arxiv.org/abs/2601.13563

作者：Aryan Karmore

【20】Patterning: The Dual of Interpretability
标题：图案化：可解释性的双重性
链接：https://arxiv.org/abs/2601.13548

作者：George Wang,Daniel Murfet

【21】Eliciting Harmful Capabilities by Fine-Tuning On Safeguarded Outputs
标题：通过对受保护输出进行微调引发有害功能
链接：https://arxiv.org/abs/2601.13528

作者：Jackson Kaunismaa,Avery Griffin,John Hughes,Christina Q. Knight,Mrinank Sharma,Erik Jones

【22】StoTAM: Stochastic Alternating Minimization for Tucker-Structured Tensor Sensing
标题：StoLAM：塔克结构张量传感的随机交替最小化
链接：https://arxiv.org/abs/2601.13522

作者：Shuang Li

【23】Preconditioning Benefits of Spectral Orthogonalization in Muon
标题：μ子光谱电离化的预处理好处
链接：https://arxiv.org/abs/2601.13474

作者：Jianhao Ma,Yu Huang,Yuejie Chi,Yuxin Chen

【24】CooperBench: Why Coding Agents Cannot be Your Teammates Yet
标题：CooperBench：为什么编码代理还不能成为你的队友
链接：https://arxiv.org/abs/2601.13295

作者：Arpandeep Khatua,Hao Zhu,Peter Tran,Arya Prabhudesai,Frederic Sadrieh,Johann K. Lieberwirth,Xinkai Yu,Yicheng Fu,Michael J. Ryan,Jiaxin Pei,Diyi Yang
备注：https://cooperbench.com

【25】The Tag is the Signal: URL-Agnostic Credibility Scoring for Messages on Telegram
标题：标签就是信号：Telegram上消息的URL不可知可信度评分
链接：https://arxiv.org/abs/2601.13294

作者：Yipeng Wang,Huy Gia Han Vu,Mohit Singhal

【26】RubRIX: Rubric-Driven Risk Mitigation in Caregiver-AI Interactions
标题：RubRIX：护理人员与人工智能互动中的文字驱动风险缓解
链接：https://arxiv.org/abs/2601.13235

作者：Drishti Goel,Jeongah Lee,Qiuyue Joy Zhong,Violeta J. Rodriguez,Daniel S. Brown,Ravi Karkar,Dong Whi Yoo,Koustuv Saha

【27】METIS: Mentoring Engine for Thoughtful Inquiry & Solutions
标题：METIS：富有洞察力的询问和解决方案的指导引擎
链接：https://arxiv.org/abs/2601.13075

作者：Abhinav Rajeev Kumar,Dhruv Trehan,Paras Chopra
备注：12 pages, 5 figures, 4 tables

【28】TinyML-Enabled IoT for Sustainable Precision Irrigation
标题：支持TinyML的物联网可持续精准灌溉
链接：https://arxiv.org/abs/2601.13054

作者：Kamogelo Taueatsoala,Caitlyn Daniels,Angelina J. Ramsunar,Petrus Bronkhorst,Absalom E. Ezugwu

【29】AI-generated data contamination erodes pathological variability and diagnostic reliability
标题：人工智能生成的数据污染侵蚀了病理变异性和诊断可靠性
链接：https://arxiv.org/abs/2601.12946

作者：Hongyu He,Shaowen Xiang,Ye Zhang,Yingtao Zhu,Jin Zhang,Hao Deng,Emily Alsentzer,Qingyu Chen,Kun-Hsing Yu,Andrew Marmenshall,Tingting Chen,Srinivas Anumasa,Daniel Ebner,Dean Ho,Kee Yuan Ngiam,Ching-Yu Cheng,Dianbo Liu
备注：*Corresponding author: Dianbo Liu (dianbo@nus.edu.sg)

【30】Your Privacy Depends on Others: Collusion Vulnerabilities in Individual Differential Privacy
标题：您的隐私取决于他人：个人差异隐私中的共谋漏洞
链接：https://arxiv.org/abs/2601.12922

作者：Johannes Kaiser,Alexander Ziller,Eleni Triantafillou,Daniel Rückert,Georgios Kaissis

【31】Actionable Interpretability Must Be Defined in Terms of Symmetries
标题：可操作的解释性必须用对称性来定义
链接：https://arxiv.org/abs/2601.12913

作者：Pietro Barbiero,Mateo Espinosa Zarlenga,Francesco Giannini,Alberto Termine,Filippo Bonchi,Mateja Jamnik,Giuseppe Marra

【32】VISPA: Pluralistic Alignment via Automatic Value Selection and Activation
标题：VISPA：通过自动价值选择和激活实现多元化调整
链接：https://arxiv.org/abs/2601.12758

作者：Shenyan Zheng,Jiayou Zhong,Anudeex Shetty,Heng Ji,Preslav Nakov,Usman Naseem
备注：WIP

【33】Neurosymbolic LoRA: Why and When to Tune Weights vs. Rewrite Prompts
标题：神经符号LoRA：为什么以及何时调整权重与重写权重
链接：https://arxiv.org/abs/2601.12711

作者：Kevin Wang,Neel P. Bhatt,Cong Liu,Junbo Li,Runjin Chen,Yihan Xi,Timothy Barclay,Alvaro Velasquez,Ufuk Topcu,Zhangyang Wang

【34】Decoding Rewards in Competitive Games: Inverse Game Theory with Entropy Regularization
标题：竞争性游戏中的奖励解码：具有熵规则化的逆博弈论
链接：https://arxiv.org/abs/2601.12707

作者：Junyi Liao,Zihan Zhu,Ethan Fang,Zhuoran Yang,Vahid Tarokh
备注：Extended journal version of ICML 2025 paper. Submitted to Operations Research

【35】Resource-Conscious RL Algorithms for Deep Brain Stimulation
标题：用于深层脑刺激的资源意识RL算法
链接：https://arxiv.org/abs/2601.12699

作者：Arkaprava Gupta,Nicholas Carter,William Zellers,Prateek Ganguli,Benedikt Dietrich,Vibhor Krishna,Parasara Sridhar Duggirala,Samarjit Chakraborty

【36】Explanation Multiplicity in SHAP: Characterization and Assessment
标题：解释SHAP中的多重性：描述和评估
链接：https://arxiv.org/abs/2601.12654

作者：Hyunseung Hwang,Seungeun Lee,Lucas Rosenblatt,Julia Stoyanovich,Steven Euijong Whang

【37】Objective Matters: Fine-Tuning Objectives Shape Safety, Robustness, and Persona Drift
标题：目标事项：微调目标塑造安全性、鲁棒性和角色漂移
链接：https://arxiv.org/abs/2601.12639

作者：Daniel Vennemeyer,Punya Syon Pandey,Phan Anh Duong,Michael Umeokoli,Samuel Ratnam

【38】What Trace Powers Reveal About Log-Determinants: Closed-Form Estimators, Certificates, and Failure Modes
标题：追踪权力揭示了关于日志决定因素的哪些信息：封闭式估计器、证书和故障模式
链接：https://arxiv.org/abs/2601.12612

作者：Piyush Sao

【39】HERMES: A Unified Open-Source Framework for Realtime Multimodal Physiological Sensing, Edge AI, and Intervention in Closed-Loop Smart Healthcare Applications
标题：赫尔姆斯：闭环智能医疗应用中的实时多模式生理传感、边缘人工智能和干预的统一开源框架
链接：https://arxiv.org/abs/2601.12610

作者：Maxim Yudayev,Juha Carlon,Diwas Lamsal,Vayalet Stefanova,Benjamin Filtjens
备注：Submitted to ACM SenSys '26, 12 pages (excl. references), 9 figures

【40】Press Start to Charge: Videogaming the Online Centralized Charging Scheduling Problem
标题：按开始充电：在线集中充电安排问题视频游戏
链接：https://arxiv.org/abs/2601.12543

作者：Alireza Ghahtarani,Martin Cousineau,Amir-massoud Farahmand,Jorge E. Mendoza
备注：41 pages

【41】Approximating splits for decision trees quickly in sparse data streams
标题：在稀疏数据流中快速逼近决策树的分裂
链接：https://arxiv.org/abs/2601.12525

作者：Nikolaj Tatti

【42】Improved Bug Localization with AI Agents Leveraging Hypothesis and Dynamic Cognition
标题：利用人工智能代理利用假设和动态认知改进漏洞定位
链接：https://arxiv.org/abs/2601.12522

作者：Asif Mohammed Samir,Mohammad Masudur Rahman
备注：13 pages, 7 tables, 5 figures

【43】Cooperative Multi-agent RL with Communication Constraints
标题：具有通信约束的协作多代理RL
链接：https://arxiv.org/abs/2601.12518

作者：Nuoya Xiong,Aarti Singh
备注：33 pages

【44】De-Anonymization at Scale via Tournament-Style Attribution
标题：通过锦标赛风格归因进行大规模去匿名化
链接：https://arxiv.org/abs/2601.12407

作者：Lirui Zhang,Huishuai Zhang
备注：14 pages

【45】Rethinking the Value of Multi-Agent Workflow: A Strong Single Agent Baseline
标题：重新思考多代理工作流程的价值：强大的单代理基线
链接：https://arxiv.org/abs/2601.12307

作者：Jiawei Xu,Arief Koesdwiady,Sisong Bei,Yan Han,Baixiang Huang,Dakuo Wang,Yutong Chen,Zheshen Wang,Peihao Wang,Pan Li,Ying Ding

【46】One-Sided Matrix Completion from Ultra-Sparse Samples
标题：超稀疏样本的单边矩阵完成
链接：https://arxiv.org/abs/2601.12213

作者：Hongyang R. Zhang,Zhenshuo Zhang,Huy L. Nguyen,Guanghui Lan
备注：41 pages

【47】EMoE: Eigenbasis-Guided Routing for Mixture-of-Experts
标题：EMoE：混合专家的特征基引导路由
链接：https://arxiv.org/abs/2601.12137

作者：Anzhe Cheng,Shukai Duan,Shixuan Li,Chenzhong Yin,Mingxi Cheng,Shahin Nazarian,Paul Thompson,Paul Bogdan
备注：accepted by ICASSP2026

【48】SynQP: A Framework and Metrics for Evaluating the Quality and Privacy Risk of Synthetic Data
标题：SynQP：用于评估合成数据质量和隐私风险的框架和工作表
链接：https://arxiv.org/abs/2601.12124

作者：Bing Hu,Yixin Li,Asma Bahamyirou,Helen Chen
备注：7 Pages, 22nd Annual International Conference on Privacy, Security, and Trust (PST2025), Fredericton, Canada

【49】Speaking to Silicon: Neural Communication with Bitcoin Mining ASICs
标题：与硅对话：与比特币采矿ASIC的神经通信
链接：https://arxiv.org/abs/2601.12032

作者：Francisco Angulo de Lafuente,Vladimir Veselov,Richard Goodman
备注：13 pages, 6 figures, 15 tables. Machine-checked Lean 4 proofs available at https://github.com/Abraxas1010/speaking-to-silicon. Validated across Antminer S9, Lucky Miner LV06, and Goldshell LB-Box platforms

【50】Why Loss Re-weighting Works If You Stop Early: Training Dynamics of Unconstrained Features
标题：为什么如果提前停止，损失重新加权有效：训练无约束特征的动态
链接：https://arxiv.org/abs/2601.12011

作者：Yize Zhao,Christos Thrampoulidis

【51】Communication-Corruption Coupling and Verification in Cooperative Multi-Objective Bandits
标题：合作多目标盗贼中的沟通-腐败耦合与验证
链接：https://arxiv.org/abs/2601.11924

作者：Ming Shi

【52】From Relative Entropy to Minimax: A Unified Framework for Coverage in MDPs
标题：从相对熵到极小极大：MDPs覆盖的统一框架
链接：https://arxiv.org/abs/2601.11890

作者：Xihe Gu,Urbashi Mitra,Tara Javidi

【53】MixFlow: Mixture-Conditioned Flow Matching for Out-of-Distribution Generalization
标题：MixFlow：用于分布外泛化的混合条件流匹配
链接：https://arxiv.org/abs/2601.11827

作者：Andrea Rubbi,Amir Akbarnejad,Mohammad Vali Sanian,Aryan Yazdan Parast,Hesam Asadollahzadeh,Arian Amani,Naveed Akhtar,Sarah Cooper,Andrew Bassett,Pietro Liò,Lassi Paavolainen,Sattar Vakili,Mo Lotfollahi

【54】RAPID-Serve: Resource-efficient and Accelerated P/D Intra-GPU Disaggregation
标题：RAPID-Serve：资源高效且加速的图形内P/D分解
链接：https://arxiv.org/abs/2601.11822

作者：Amna Masood,Pratishtha Gaur,Nuwan Jayasena

【55】A Proof of Concept for a Digital Twin of an Ultrasonic Fermentation System
标题：超声波发酵系统数字双胞胎的概念验证
链接：https://arxiv.org/abs/2601.11723

作者：Francesco Saverio Sconocchia Pisoni,Andrea Vitaletti,Davide Appolloni,Federico Ortenzi,Blasco Morozzo della Rocca,Mariano José Guillén,Alessandro Contaldo
备注：23 pages, submitted to the 22nd International Conference on Intelligent Environments (IE 2026)

【56】Telling Human and Machine Handwriting Apart
标题：区分人类和机器手写
链接：https://arxiv.org/abs/2601.11700

作者：Luis A. Leiva,Moises Diaz,Nuwan T. Attygalle,Miguel A. Ferrer,Rejean Plamondon

【57】Activation Sensitivity as a Unifying Principle for Post-Training Quantization
标题：激活敏感性作为训练后量化的统一原则
链接：https://arxiv.org/abs/2601.11663

作者：Bruce Changlong Xu

【58】Zeros can be Informative: Masked Binary U-Net for Image Segmentation on Tensor Cores
标题：零可以是信息性的：用于张量核图像分割的掩蔽二进制U-Net
链接：https://arxiv.org/abs/2601.11660

作者：Chunshu Wu,Ruibing Song,Sushant Kondguli,Tong Geng,Ang Li

【59】A Multimodal Data Processing Pipeline for MIMIC-IV Dataset
标题：MMIC-IV数据集的多模式数据处理管道
链接：https://arxiv.org/abs/2601.11606

作者：Farzana Islam Adiba,Varsha Danduri,Fahmida Liza Piya,Ali Abbasi,Mehak Gupta,Rahmatollah Beheshti

【60】Let Me Try Again: Examining Replay Behavior by Tracing Students' Latent Problem-Solving Pathways
标题：让我再试一次：通过追踪学生潜在的问题解决途径来检查重播行为
链接：https://arxiv.org/abs/2601.11586

作者：Shan Zhang,Siddhartha Pradhan,Ji-Eun Lee,Ashish Gurung,Anthony F. Botelho
备注：16 pages, 7 figures, LAK2026

【61】Uniqueness ratio as a predictor of a privacy leakage
标题：单一比率作为隐私泄露的预测因素
链接：https://arxiv.org/abs/2601.11550

作者：Danah A. AlSalem AlKhashti

【62】Opportunities in AI/ML for the Rubin LSST Dark Energy Science Collaboration
标题：Rubin LCST暗能量科学合作组织的AI/ML机会
链接：https://arxiv.org/abs/2601.14235

作者：LSST Dark Energy Science Collaboration,Eric Aubourg,Camille Avestruz,Matthew R. Becker,Biswajit Biswas,Rahul Biswas,Boris Bolliet,Adam S. Bolton,Clecio R. Bom,Raphaël Bonnet-Guerrini,Alexandre Boucaud,Jean-Eric Campagne,Chihway Chang,Aleksandra Ćiprijanović,Johann Cohen-Tanugi,Michael W. Coughlin,John Franklin Crenshaw,Juan C. Cuevas-Tello,Juan de Vicente,Seth W. Digel,Steven Dillmann,Mariano Javier de León Dominguez Romero,Alex Drlica-Wagner,Sydney Erickson,Alexander T. Gagliano,Christos Georgiou,Aritra Ghosh,Matthew Grayling,Kirill A. Grishin,Alan Heavens,Lindsay R. House,Mustapha Ishak,Wassim Kabalan,Arun Kannawadi,François Lanusse,C. Danielle Leonard,Pierre-François Léget,Michelle Lochner,Yao-Yuan Mao,Peter Melchior,Grant Merz,Martin Millon,Anais Möller,Gautham Narayan,Yuuki Omori,Hiranya Peiris,Laurence Perreault-Levasseur,Andrés A. Plazas Malagón,Nesar Ramachandra,Benjamin Remy,Cécile Roucelle,Jaime Ruiz-Zapatero,Stefan Schuldt,Ignacio Sevilla-Noarbe,Ved G. Shah,Tjitske Starkenburg,Stephen Thorp,Laura Toribio San Cipriano,Tilman Tröster,Roberto Trotta,Padma Venkatraman,Amanda Wasserman,Tim White,Justine Zeghal,Tianqing Zhang,Yuanyuan Zhang
备注：84 pages. This is v1.0 of the DESC's white paper on AI/ML, a collaboration document that is being made public but which is not planned for submission to a journal

【63】SCG With Your Phone: Diagnosis of Rhythmic Spectrum Disorders in Field Conditions
标题：SCG与您的手机：现场条件下节律谱障碍的诊断
链接：https://arxiv.org/abs/2601.13926

作者：Peter Golenderov,Yaroslav Matushenko,Anastasia Tushina,Michal Barodkin

【64】Distribution-Free Confidence Ellipsoids for Ridge Regression with PAC Bounds
标题：具有PAC界的岭回归的无分布置信椭圆体
链接：https://arxiv.org/abs/2601.13436

作者：Szabolcs Szentpéteri,Balázs Csanád Csáji

【65】Scaling laws for amplitude surrogates
标题：幅度替代物的标度定律
链接：https://arxiv.org/abs/2601.13308

作者：Henning Bahl,Victor Bresó-Pla,Anja Butter,Joaquín Iturriza Ramirez
备注：45 pages, 20 figures

【66】Empirical Risk Minimization with $f$-Divergence Regularization
标题：通过$f$实现经验风险最小化-分歧监管
链接：https://arxiv.org/abs/2601.13191

作者：Francisco Daunas,Iñaki Esnaola,Samir M. Perlaza,H. Vincent Poor
备注：Submitted to IEEE Transactions on Information Theory. arXiv admin note: substantial text overlap with arXiv:2502.14544, arXiv:2508.03314

【67】Reorienting off-path Nudged Elastic Bands (RONEB) via Minimum Mode Following
标题：通过最小模式跟随重新定向偏离路径轻推弹性带（RONEB）
链接：https://arxiv.org/abs/2601.12630

作者：Rohit Goswami,Miha Gunde,Hannes Jónsson
备注：25 pages. 11 figures

【68】Deterministic and probabilistic neural surrogates of global hybrid-Vlasov simulations
标题：全球混合Vlasov模拟的确定性和概率神经代理
链接：https://arxiv.org/abs/2601.12614

作者：Daniel Holmberg,Ivan Zaitsev,Markku Alho,Ioanna Bouri,Fanni Franssila,Haewon Jeong,Minna Palmroth,Teemu Roos

【69】Temporal Data and Short-Time Averages Improve Multiphase Mass Flow Metering
标题：时间数据和短期动态改善多相质量流量计量
链接：https://arxiv.org/abs/2601.12433

作者：Amanda Nyholm,Yessica Arellano,Jinyu Liu,Damian Krakowiak,Pierluigi Salvo Rossi
备注：9 pages, 6 figures

【70】AQUA-Bench: Beyond Finding Answers to Knowing When There Are None in Audio Question Answering
标题：AQUA-Bench：超越在音频问题回答中寻找知道何时没有答案
链接：https://arxiv.org/abs/2601.12248

作者：Chun-Yi Kuan,Hung-yi Lee
备注：Accepted to ICASSP 2026. Project Website: https://kuan2jiu99.github.io/AQUA-Bench-demo/

【71】Inter-Cell Interference Rejection Based on Ultrawideband Walsh-Domain Wireless Autoencoding
标题：基于超宽带沃尔什域无线自编码的小区间干扰抑制
链接：https://arxiv.org/abs/2601.11713

作者：Rodney Martinez Alonso,Cel Thys,Cedric Dehos,Yuneisy Esthela Garcia Guzman,Sofie Pollin
备注：This preprint was submitted to The 2026 EuCNC & 6G Summit

【72】Anisotropic Tensor Deconvolution of Hyperspectral Images
标题：超光谱图像的各向异性张量反卷积
链接：https://arxiv.org/abs/2601.11694

作者：Xinjue Wang,Xiuheng Wang,Esa Ollila,Sergiy A. Vorobyov
备注：To appear in ICASSP 2026

【73】AI Agents Need Memory Control Over More Context
标题：人工智能代理需要对更多上下文进行记忆控制
链接：https://arxiv.org/abs/2601.11653

作者：Fouad Bousetouane
备注：32 pages, 7 figures

【74】Multi-Scale Negative Coupled Information Systems (MNCIS): A Unified Spectral Topology Framework for Stability in Turbulence, AI, and Biology
标题：多尺度负耦合信息系统（MNCIS）：湍流、人工智能和生物稳定性的统一谱布局框架
链接：https://arxiv.org/abs/2601.11594

作者：Pengyue Hou
备注：Includes supplementary materials and code. Foundation and mathematical proofs can be found in the companion paper arXiv:2601.00638

机器翻译由腾讯交互翻译提供，仅供参考

点击“阅读原文”获取带摘要的学术速递