2026-11-20 | CS.LG机器学习 | 共 119 篇
机构等信息由AI生成,仅供参考
快速导航
1. 深度学习架构与训练方法 3 篇
2. 表示学习、自监督与对比学习 1 篇
3. 强化学习与序列决策 6 篇
4. 生成模型与概率建模 1 篇
5. 优化、泛化与理论分析 2 篇
6. 高效学习、压缩与部署 1 篇
7. 联邦学习、隐私与安全 2 篇
8. 鲁棒性、不确定性与可信学习 4 篇
9. 图学习与结构化数据 1 篇
10. 迁移、元学习与持续学习 3 篇
11. 数据集、基准与评测 3 篇
12. 机器学习应用 4 篇
13. 其他/综合机器学习 9 篇
1. 深度学习架构与训练方法 | 3 篇
1. DS3M: Deep structured state space models for nonlinear system identification on modern hardware
DS3M:面向现代硬件的非线性系统辨识深度结构化状态空间模型
AI 总结:针对SUBNET训练速度慢的问题,提出DS3M模型,结合Triton自定义GPU内核降低开销,通过基准比较验证其在精度和训练时间上的优势。
链接:https://arxiv.org/abs/2611.25274
机构:University of Wuppertal(伍珀塔尔大学)
作者:Marco Roschkowski, Hannes Gernandt
英文摘要:Neural state-space identification learns dynamical models from measured inputs and outputs. SUBNET combines this task with an encoder that estimates initial states from short measurement histories. Its continuous-time version requires many successive neural-network evaluations during numerical integration, resulting in slow training speed which is a major limitation for practical use cases. We propose deep structured state space models (DS3M), which combine compact nonlinear dynamical components in parallel and in series. Custom GPU kernels implemented with Triton reduce the launch and memory overhead of their sequential evaluation. We explain this construction, describe a port-Hamiltonian variant, and present various benchmark comparisons of accuracy and training time.
2. Protein Functional Semantic Alignment via Representation Rectification
基于表示校正的蛋白质功能语义对齐
AI 总结:该研究针对蛋白质语言模型表示中功能信号与非功能变异纠缠的问题,提出SAVR对比式后训练方法,利用Pfam家族共成员关系重塑表示,在EC基准达最优、GO基准具竞争力,可改进蛋白质功能预测。
链接:https://arxiv.org/abs/2611.25396
机构:The University of Sydney(悉尼大学); Shanghai Jiao Tong University(上海交通大学)
作者:Chuyang Zhou, Bingxin Zhou, Andi Han, Chang Xu
英文摘要:Protein language models (PLMs) are commonly pretrained with likelihood-based objectives such as masked language modeling or denoising. These objectives capture sequence-level regularities, but can produce representations in which functional signals are entangled with non-functional variation (e.g., sequence identity, compositional bias, and frequent local motifs), creating a readout bottleneck for downstream function prediction. We introduce Functional Semantic Alignment via Representation Rectification (SAVR), a contrastive post-training method using Pfam family co-membership as a homology-guided discrete semantic signal to reshape PLM representations. SAVR reduces directional dominance and redistributes variance across embedding dimensions, decreasing variance concentration, and increasing effective dimensionality of the embedding covariance spectrum. Without observing any task-specific labels during post-training, SAVR achieves state-of-the-art results on Enzyme Commission (EC) benchmarks and competitive results on Gene Ontology (GO) benchmarks under frozen-backbone evaluation. These results suggest that homology-guided contrastive post-training can rectify PLM representation geometry and improve sequence-based protein function prediction.
3. Interference Bounds and Merge-Aware Training Dynamics for Composable Fine-Tuned Models
可组合微调模型的干扰界与感知合并的训练动力学
AI 总结:该研究提出可组合微调模型的干扰界,证明合并模型超额损失由干扰泛函控制,提出曲率引导正则化器,实验验证其效果并对比各向同性惩罚。
链接:https://arxiv.org/abs/2611.25430
机构:Swinburne University of Technology(斯威本科技大学)
作者:Alireza Moayedikia
英文摘要: Combining fine-tuned models by composing their parameter updates has become a standard alternative to multi-task training. We formalize composability as a property of a collection of task vectors and prove that the excess loss of the merged model is controlled by an interference functional of the cross-task curvature overlap and the signed alignment of the updates. In the quadratic regime this functional equals the excess loss, giving a criterion that is necessary as well as sufficient and that measures the angle between task vectors in the curvature metric rather than the Euclidean one. Fine-tuning under a curvature-guided regularizer -- constructed from curvature statistics of the kind memory-efficient optimizers maintain -- keeps displacement in the base curvature metric below a ceiling set by the penalty coefficient at every iterate, at the standard $O(1/\sqrt{T})$ non-convex rate. Under a single Loewner comparison between the base statistic and the task curvatures, this control converts into a ceiling on the interference functional, inversely proportional to the penalty coefficient, so that the curvature term of the bound is controlled during training. A first-order term, present because regularized solutions are not stationary for the task loss, is not reduced in the same way. Experiments on an exact non-quadratic model confirm the budget and the ceiling, show the first-order term dominating for strong penalties, and find the curvature-guided penalty slightly better than an isotropic one. In a three-seed transformer study the functional is rank-correlated with merging degradation, but no more strongly than the Euclidean displacement norm.
2. 表示学习、自监督与对比学习 | 1 篇
4. Separating Interpretability and Reconstruction in Sparse Autoencoders
分离稀疏自编码器中的可解释性与重构
AI 总结:研究针对传统稀疏自编码器的重构保真度与可解释性难对齐问题,提出模块化框架DualSAE,通过两阶段训练与去相关损失实现二者分离,在多数据集上获显著提升且下游应用表现更好。
链接:https://arxiv.org/abs/2611.25165
机构:Stony Brook University(石溪大学); Concordia University(康考迪亚大学); Mila–Quebec AI Institute(米拉-魁北克人工智能研究所)
作者:Anh Tien Nguyen, Dimitris Samaras, Mahdi S. Hosseini
英文摘要:Sparse Autoencoders (SAEs) are effective tools for mechanistic interpretability by decomposing neural representations into sparse and interpretable features. However, a single sparse representation in traditional SAEs is responsible for both reconstruction fidelity and interpretability, which are not guaranteed to align. To separate these objectives, we propose DualSAE, a modular framework that augments standard SAEs with a plug-in Residual Branch. A standard SAE serves as the Sparse Branch, learning a sparse code that contains the dominant semantics of the input. The added Residual Branch models the non-semantic information omitted by the Sparse Branch. We ensure that the contributions of the two branches towards the overall reconstruction have minimal overlap through a decorrelation loss. We also adopt a two-stage training strategy that first stabilizes sparse representation learning before coupling with the Residual Branch. Across diverse encoder-SAE configurations on two vision datasets and two language corpora, DualSAE achieves mean relative improvements of approximately 29% in reconstruction fidelity and 34% in interpretability (measured by monosemanticity score). Moreover, with more interpretable representations, DualSAE yields consistent performance gains in the downstream applications of topic modeling, hypothesis generation, and LLM steering. Qualitative analysis further suggests that the Sparse Branch indeed preserves dominant semantic content, whereas the Residual Branch captures non-semantic information to improve reconstruction fidelity.
3. 强化学习与序列决策 | 6 篇
5. Bayesian $κ$-Explorer for Minimax Model Estimation in MDPs
用于马尔可夫决策过程(MDP)中极小极大模型估计的贝叶斯κ-探索器
AI 总结:针对MDP中极小极大转移估计问题,提出贝叶斯κ-探索器,通过狄利克雷先验简化流程,保留收敛率且在多个基准任务上实现更低的最坏情况转移估计误差。
链接:https://arxiv.org/abs/2611.25160
机构:University of California, San Diego(加利福尼亚大学圣地亚哥分校); University of Southern California(南加州大学)
作者:Xihe Gu, Urbashi Mitra, Tara Javidi
英文摘要:We study active exploration for minimax transition estimation in unknown Markov decision processes (MDPs). Frequentist $\kappa$-Explorer~\citep{gu2026kappa} addresses this problem through a parameterized family of exploration objectives, but requires confidence-bound constructions and joint optimization over plausible transition kernels and occupancy measures, introducing a computational burden that limits its practical applicability. We develop Bayesian $\kappa$-Explorer, which simplifies this procedure by placing Dirichlet priors on the transition distributions. Under the posterior mean estimator, the posterior expected squared $\ell_2$ error equals the total posterior variance, providing a closed-form exploration signal. The algorithm plans under the posterior mean model, eliminating confidence-set optimization. Under the stated assumptions, this simpler planning procedure retains the $\widetilde{O}(T^{-1/3})$ surrogate convergence rate of frequentist $\kappa$-Explorer. We further bound the maximum posterior variance relative to the optimal asymptotic minimax allocation benchmark. Experiments on Pendulum, Mountain Car, and Taxi demonstrate lower worst-case transition-estimation errors than the tested baselines.
6. ϕTD: Distributional Reinforcement Learning using Characteristic Functions
ϕTD:使用特征函数的分布强化学习
AI 总结:本研究提出φTD方法,通过频域计算损失突破分布强化学习的分布类别局限,统一分位数与分类方法,在合成基准及大规模深度RL环境中验证其灵活性可提升性能。
链接:https://arxiv.org/abs/2611.25208
机构:Polytechnique Montréal(蒙特利尔理工学院); University of Toronto(多伦多大学); Mila – Quebec Artificial Intelligence Institute(米拉-魁北克人工智能研究所); Vector Institute(矢量研究所)
作者:Thomas Mousseau, Tyler Kastner, Davide Baldelli, Amir-massoud Farahmand
英文摘要: Distributional Reinforcement Learning methods aim to learn the entire distribution of returns, yet current algorithmic approaches are limited to a narrow class of distributions. We propose $\varphi$TD, a method to overcome this limitation by computing the loss in the frequency domain. We demonstrate that this further unifies quantile and categorical approaches to distributional RL, and enables the learning of distributions that were previously impossible to capture, such as mixtures of continuous distributions. We empirically study $\varphi$TD, across synthetic benchmarks and large-scale deep RL environments and demonstrate that the flexibility obtained from wider parametric families can lead to improved performance.
7. Cross-Question Reliable Reinforcement Learning
跨问题可靠强化学习
AI 总结:该研究针对AI置信度估计的缺陷,提出CARE方法,通过跨问答对计算奖励强化置信度,在数学及多类分布外基准上显著提升了5%风险下的问题覆盖度。
链接:https://arxiv.org/abs/2611.25292
机构:TU Darmstadt(达姆施塔特工业大学)
作者:Hector G. Rodriguez, Marcus Rohrbach
英文摘要:AI systems operating near the limits of their capabilities can maximize utility while minimizing the risk of error by abstaining or deferring questions they are not confident about to a human or a stronger model. Recently, large language models (LLMs) have been post-trained to estimate confidence in their own answers. To achieve this, methods reinforce confidence scores based on the correctness signal to each question-answer pair individually, with rewards that encourage assigning confidence 1 to correct answers and 0 to incorrect answers. However, estimating confidence scores is useful in that it enables selecting only the most confident answers, to maximize the share of solved questions while accommodating the user's risk tolerance. Assigning confidence 1 to correct answers and 0 to incorrect ones is theoretically optimal, but unattainable in practice and detrimental for selective prediction with a non-zero risk tolerance: confidences pushed to 0 and 1 carry no information about how answers compare to one another. We therefore introduce CARE (Cross-question Abstention Reward Estimation), which computes rewards over a set of question-answer pairs: At each of several risk tolerances, the confidences across the set induce an abstention threshold, and each confidence is rewarded for lying on the correct side of it. Since the threshold itself comes from the set, every verbalized confidence is reinforced relative to other questions' confidence, not against its own binary correctness label. On math, CARE increases the coverage at 5% risk (C@5) across MATH-500, GSM8K, and Big-Math-Digits from 0.8% to 42.4%, compared to existing pointwise correctness-matching rewards. When trained on multi-hop Wikipedia QA from HotpotQA, CARE improves C@5 in out-of-distribution general-knowledge, reasoning, and math benchmarks (including CommonsenseQA, GPQA, SimpleQA, and TriviaQA) from 12.3% to 23.6%.
8. Improving Imitation Policies for Multi-Phase Tasks through Residual Learning Allocation
通过残差学习分配改进多阶段任务的模仿策略
AI 总结:该研究针对多阶段任务的模仿策略,提出基于完成增益分配残差强化学习的训练预算,在机器人对接与装配实验中验证了该方法可提升任务完成度。
链接:https://arxiv.org/abs/2611.25320
机构:Institute for Systems Research, University of Maryland(马里兰大学系统研究所)
作者:Clinton Enwerem, Saleh Kemal, John S. Baras, Calin Belta
英文摘要:Residual reinforcement learning corrects an imitation-trained base policy using a limited budget of environment interaction. In a task with several phases, this budget must be divided among the phases, and the phase in which the base policy fails is not necessarily the phase in which a correction most improves task completion. We formalize this distinction through the completion gain: the change in full-task completion probability when, during a single phase, a state-feedback controller replaces selected components of the base policy's action. We decompose the completion gain into the probability of reaching the phase and the conditional improvement in completion from the start of the phase, and we estimate it from paired episodes. The positive estimates determine which action components the residual policy corrects and how training episodes are distributed across phases. In stochastic point-mass docking experiments, most failures occurred in the final docking phase, whereas the largest completion gain occurred in the preceding channel phase; gain-based allocation improved completion over allocation proportional to failure counts and, in a variant with a longer approach, over uniform allocation at half of the budget. In simulated bimanual robotic assembly, the completion-gain estimates were positive only for insertion and release, although few failures occurred during release; in a second assembly configuration, a residual policy trained in the release phase improved completion from held-out states at the start of release by 16.9%.
9. Drifting, Discriminating, and Directioning for Action Coordination in Higher-dimensional Action Space
针对高维动作空间中动作协调的漂移、判别与定向方法
AI 总结:本文提出轻量级漂移策略模型及三步学习算法,通过分布协同优化、对抗判别与N选优采样,提升高维动作空间强化学习的样本效率与控制性能。
链接:https://arxiv.org/abs/2611.25385
机构:National Institute of Informatics(情报通信研究机构); The Graduate University for Advanced Studies, SOKENDAI(总合研究大学院大学(SOKENDAI))
作者:Taisuke Kobayashi
英文摘要:This paper proposes a lightweight policy model to mitigate the difficulty of handling higher-dimensional action spaces in reinforcement learning (RL), alongside an algorithm to effectively train it. To address this, we focus on drifting models, which can implicitly represent complex distributions and infer actions in a single forward pass. However, while drifting models can be straightforwardly trained via backpropagation of the Q-gradient, this alone cannot be expected to appropriately capture the desired action coordination. Therefore, this paper constructs a three-step learning algorithm to encourage the action coordination. First, the source distribution of the drifting model is co-optimized as a diagonal policy model. This conditions the drifting model to attach the coordination, which cannot be expressed by a diagonal distribution, as residuals. Second, a discriminator model before and after drifting is introduced with adversarial training for a regularization term. This actively promotes the extraction of the coordination that fundamentally separates the two distributions. Third, best-of-N sampling is applied for action selection during rollouts in order to give directionality toward/beyond the regions where the coordination is insufficiently extracted/sufficiently discovered. The proposed method consistently improves the sample efficiency and the final control performance over the conventional diagonal distribution modeling across three locomotion tasks with high-dimensional action spaces. Furthermore, the analysis indicates that this improvement is achieved by reducing variance in certain action dimensions, effectively coupling them with other actions as action coordination.
10. Coordinate Basis Linear Bandits with Sublinear Regret over Low-Rank Clusters
低秩聚类下具有次线性遗憾的坐标基线性老虎机
AI 总结:针对有限臂集高维线性老虎机的计算效率问题,提出CBL算法,通过聚类机制与正交基保留协方差信息,将每轮计算量降至$O(\bar r^2)$,实现鲁棒次线性遗憾。
链接:https://arxiv.org/abs/2611.25424
机构:University of Science and Technology of China(中国科学技术大学)
作者:Hantao Yang, Hong Xie, Defu Lian
英文摘要:We study the problem of computational efficiency for linear bandits in high-dimensional settings with a finite arm set. In linear bandits, the increase in the dimension $d$ of the feature vectors leads to growing computational costs of $O(d^2)$ at each round of update. Traditional low-rank approximation methods such as SOFUL reduce computation via fixed-size matrix sketching, yet run the risk of incurring vacuous linear regret when the spectral tail of the data is heavy and the approximation rank is inadequately selected. To guarantee regret convergence and effectively reduce computational costs, we introduce a clustering mechanism and propose the Coordinate Basis Linear Bandit (CBL) algorithm. Our method preserves the full covariance information in each cluster by maintaining an exact orthonormal basis of the cluster's column space together with a low-dimensional coordinate representation, which guarantees robust sublinear regret without spectral-tail vulnerabilities. It performs cluster switching by assigning a sentinel for each cluster, and reduces per-round update computation to $O(\bar r^2)$ in the cluster's effective dimension $\bar r=\min(r,l)$, where $r$ is the intrinsic rank of the data and $l
4. 生成模型与概率建模 | 1 篇
11. Process-Knowledge-Guided Latent Diffusion for Generative Design of Porous Architectures: A Case Study in Aerosol Jet Printing
工艺知识引导的潜在扩散模型用于多孔结构的生成式设计:以气溶胶喷射打印为例
AI 总结:该研究提出工艺知识引导的潜在扩散模型,结合VQ-VAE、拓扑U-Net、ControlNet及回归器,在7种气溶胶喷射打印流量参数下,可生成适配特定工艺的多孔布局,评分随权重变化表现良好。
链接:https://arxiv.org/abs/2611.25264
机构:Arizona State University(亚利桑那州立大学)
作者:Ngoc Dang Khoa Nguyen, Hyunwoong Ko
英文摘要:Printed porous layouts interleave functional material with open channels, and the layout that uses the material best depends on how the process settings shape the deposit. Generative models propose layouts quickly, but a generator that does not receive the process setting samples from the same distribution at every setting, even when the process-dependent score ranks candidate layouts differently. This paper examines whether process knowledge can steer a latent diffusion generator toward layouts suited to a given process setting in a prescribed design domain. We formulate a process-dependent utilization score in which a process model sets the length over which the utility of printed material decays with distance from a channel. The generator combines a vector-quantized variational autoencoder (VQ-VAE) latent representation, a topology U-Net, a ControlNet domain branch, and a separately trained regressor that predicts the score from noised latents; the domain branch conditions each denoising step and the regressor gradient steers it. The method is applied to aerosol jet printing, where the atomizer and sheath gas flows enter the score through the flow-focusing relation. Across seven modeled flow settings, the lowest-scoring candidate layout of a domain differed between settings in 191 of 500 domains. In a sampling study, the mean directly computed score on three validation domains and one test domain was 0.642 with domain conditioning alone, 0.636 at a weight of 300, and 0.450 at 12,000 (112 scores per weight; lower is better). The results support process-knowledge-guided generation as an approach to process-specific porous design under the stated assumptions.
5. 优化、泛化与理论分析 | 2 篇
12. Near-Quadratic Sample Compression for All VC Classes
所有VC类的近二次样本压缩
AI 总结:该研究针对学习理论中的样本压缩猜想,构造了大小为O(d² log² d)的样本压缩方案,证明了每个VC维为d的二元概念类存在d的多项式大小的样本压缩方案这一较弱猜想。
链接:https://arxiv.org/abs/2611.25011
机构:Technion(以色列理工学院); Yale University(耶鲁大学); Google(谷歌)
作者:Yang Cai, Vineet Gupta, Yanchen Jiang, Christopher Liaw, Aranyak Mehta, Shay Moran, Grigoris Velegkas, Di Wang, Manfred K. Warmuth
英文摘要:The sample compression conjecture is a longstanding open problem in learning theory, originating in the work of Littlestone and Warmuth in the 1980s. At its heart is a connection between simplification and generalization: if a predictor consistent with a training sample is defined using only a small part of that sample, then it is likely to generalize. Sample compression schemes formalize this idea, and many natural learning algorithms can be analyzed this way. The sample compression conjecture of Floyd and Warmuth (1995) asserts that every binary concept class of VC dimension $d$ admits a sample compression scheme of size $O(d)$; a weaker conjecture asks only for size polynomial in $d$. Both have remained open: the first bound depending only on $d$, due to Moran and Yehudayoff (2016), is exponential in $d$. We prove the weaker conjecture by constructing a sample compression scheme of size $O(d^2 \log^2 d)$.
13. HapBO: Harnessed $α$-$π$ Bayesian Optimization via Expert Natural Language Priors
HapBO:基于专家自然语言先验的受控α-π贝叶斯优化
AI 总结:HapBO是基于专家自然语言先验的贝叶斯优化方法,可从专家或LLM获取信念,在40个HPOBench任务及实际应用中表现优异,零循环内令牌成本且成本远低于同类方法。
链接:https://arxiv.org/abs/2611.25100
机构:Polytechnique Montréal(蒙特利尔理工学院); MILA – Québec AI Institute(MILA – 魁北克人工智能研究所); Université de Montréal(蒙特利尔大学)
作者:Jean-Sébastien Giroux, Rose Guay Hottin, Marco Bonizzato
英文摘要:Bayesian Optimization (BO) can be substantially more sample-efficient when informed by prior knowledge about the problem. In practice, however, such knowledge is often available only in natural language, e.g., expert intuitions or documentation, rather than as the explicit probability distribution landscape that prior-guided BO methods may require. Leveraging it effectively demands both an expressive mapping from language to priors and robustness to information that may be misleading. We propose HapBO (Harnessed $\alpha$-$\pi$ Bayesian Optimization), which elicits beliefs once from a human expert, a large language model (LLM), or both, and deterministically compiles a structured specification into a reusable belief function ($\pi: \mathcal{X} \to[0,1]$), with a rich representation not restricted to a Gaussian family. Building on $\alpha$-$\pi$BO, HapBO applies this belief function through a bounded positive multiplier on any nonnegative acquisition, preserving positive acquisition support and requiring no LLM calls during optimization. We prove that, for the Expected Improvement acquisition function (EI), HapBO preserves the standard simple-regret rate up to a bounded factor, even under adversarial beliefs. We evaluate HapBO on 40 HPOBench tasks, a synthetic suite in 6-15 dimensions and two real-world applications: cortical neurostimulation and constrained solar design. HapBO achieves the lowest aggregate HPOBench regret and strong real-world task performance, with zero in-loop token cost. In real-world applications it improves the optimization outcome even at budgets of 12-20 queries, and at a small fraction of the wall-clock and token costs of other LLM-based methods.
6. 高效学习、压缩与部署 | 1 篇
14. Learning to Verify: Signature-Guided Training for Adaptive Adversarial Detection
学习验证:用于自适应对抗检测的签名引导训练
AI 总结:该研究提出双模型方法,用激活签名引导训练验证器,在公开数据集上针对自适应攻击,验证器优于基于置信度和特征的检测器,同时发现其在跨扰动类型的鲁棒性存在局限。
链接:https://arxiv.org/abs/2611.25271
机构:University of New Brunswick(新不伦瑞克大学)
作者:Arash Vashagh, Roozbeh Razavi-Far
英文摘要:Adversarial example detectors often fail when attackers explicitly attempt to bypass them. We propose a dual-model approach that keeps the original classifier fixed and trains a separate verifier to assess its predictions. During training, we identify patterns of internal neural network activations that distinguish genuine examples of each class from adversarial examples targeting that class. These patterns, called activation signatures, guide the generation of challenging adversarial examples that mislead the classifier, challenge the verifier, and internally resemble the target class. The verifier learns to recognize these examples while preserving its performance on clean inputs through knowledge distillation. At inference, the verifier evaluates the classifier's prediction without requiring activation signatures. We evaluate the approach on publicly available datasets against several adaptive attacks, including attacks unseen during training. The proposed verifier substantially outperforms directly attacked confidence-based and feature-based detectors under the evaluated L-infinity attacks. Ablation studies show that adaptive verifier training is the main source of this improvement, while signature guidance primarily benefits clean accuracy. Additional L2 evaluations reveal limitations in robustness across perturbation types.
7. 联邦学习、隐私与安全 | 2 篇
15. Federated Learning of Car Aerodynamics Surrogates with Spectral Optimizers
基于谱优化器的汽车空气动力学代理模型联邦学习
AI 总结:本研究针对汽车空气动力学代理模型训练数据专有问题,采用联邦学习框架,对比谱与非谱优化器,发现FedMARS-M等谱优化器在低本地训练周期下阻力误差远低于非谱方法,性能更优。
链接:https://arxiv.org/abs/2611.25300
作者:Arman Bolatov, Nikolay Spitsyn, Daniil Medyakov, Vladimir Aletov, Dmitry Bylinkin, Martin Takáč, Aleksandr Beznosikov
英文摘要:Neural surrogates predict the aerodynamic forces on a car in seconds, compared with hours for a CFD simulation, but training them requires thousands of CFD simulations of different car designs, which car manufacturers keep proprietary. Federated learning allows several manufacturers to train a shared surrogate without exchanging data. We study which federated optimizer brings such a surrogate closest to one trained on pooled data. On DrivAerNet++, we train a Transolver surrogate across six clients that each hold cars of a single body type, and we compare three spectral optimizers, which orthogonalize the update of each weight matrix (FedMuon and two methods we propose, FedMARS-M and FedOLion), with three non-spectral ones (FedAvg, SCAFFOLD-Adam and FedLion), all tuned with the same protocol. FedMuon and FedMARS-M reach a drag error within 2% of centralized training with Muon, using two to four local epochs per round. The non-spectral methods perform best with sixteen local epochs, the largest number we tested, and still have a 17 to 37% higher drag error than FedMARS-M. We relate this gap to the low stable rank of the weight gradients, for which an orthogonalized step yields about twice the first-order decrease of a gradient step of the same spectral norm. Code: this https URL
16. SplitReAlign: Partition-Aware Representation Realignment for Split Federated Unlearning
SplitReAlign:面向分裂式联邦遗忘的分区感知表示重对齐
AI 总结: 针对分裂式联邦遗忘的复杂问题,提出SplitReAlign框架,通过分区感知表示重对齐技术,在CIFAR-10和CIFAR-100上取得优于对比方法的遗忘效果。
链接:https://arxiv.org/abs/2611.25301
机构:School of Computer Science and Artificial Intelligence, Wuhan Textile University(武汉纺织大学计算机科学与人工智能学院); Engineering Research Center of Hubei Province for Clothing Information, Wuhan Textile University(武汉纺织大学湖北省服装信息化工程研究中心); Australian Institute for Machine Learning (AIML), Adelaide University(阿德莱德大学澳大利亚机器学习研究所)
作者:Jia Chen, Yecheng Lv, Xinrong Hu, Jun Bai
英文摘要:Split federated learning (SFL) distributes neural networks across clients and a server, coupling class-specific knowledge across their representations. This coupling complicates split federated unlearning (SFU), where class-wise forgetting must remove target knowledge while preserving retained utility under heterogeneous representation spaces and asymmetric update constraints. We propose SplitReAlign, a partition-aware representation realignment framework for class-wise SFU. It constructs endpoint-specific discriminative subspaces coordinated through a shared semantic anchor, selectively realigning forgotten representations while penalizing changes outside the active subspaces. To accommodate asymmetric update constraints, SplitReAlign leverages gradients through frozen servers for Client-only updates and fixed client-feature statistics for Server-only guidance, while Joint updates optimize both partitions. Experiments on CIFAR-10 and CIFAR-100 under heterogeneous client distributions demonstrate that SplitReAlign achieves the lowest average gap to retraining among all compared methods across six settings. Notably, Server-only achieves the closest agreement with retraining on both datasets, highlighting the effectiveness of partition-aware guidance under restricted updates.
8. 鲁棒性、不确定性与可信学习 | 4 篇
17. Density Feedback Control for Transferability of Generative Sparse Adversarial Attacks
用于生成式稀疏对抗攻击可迁移性的密度反馈控制
AI 总结:该研究针对生成式稀疏对抗攻击的密度控制问题,提出DFC方法,通过反馈调整稀疏系数,提升了部分攻击的迁移成功率,且在更少像素扰动下达到目标密度。
链接:https://arxiv.org/abs/2611.25001
机构:Korea Aerospace University(韩国航空大学)
作者:Dongsu Song, Jay Hoon Jung
英文摘要:Generator-based sparse adversarial attacks train encoder--decoder networks to produce transferable sparse perturbations for black-box robustness evaluation. Their perturbation density, however, is set only indirectly by a fixed sparsity coefficient whose mapping to density is unknown and shifts during training, so reaching a target density requires re-tuning; this fixed pressure also yields little transferability gain once density reduction begins. We propose Density Feedback Control (DFC), which tightens or relaxes this coefficient at the end of each epoch according to the gap between the observed hard-mask density and the target, without changing the generator architecture or loss. At a 10\% target density, DFC brings the mean density of three base attacks within one percentage point of the target in a single run and improves the transfer attack success rate (ASR) of two of them on CNN and Transformer targets, including under per-image $\ell_0$ budgets, while perturbing fewer pixels. Tracing GPG's training, we observe that as DFC relaxes the coefficient, the transfer ASR recovers with the density and overtakes the fixed-coefficient baseline with fewer pixels. For EGS-TSSA, which restricts perturbations to source-model feature-importance regions, DFC yields no consistent gain; we analyze how this restriction interacts with density feedback.
18. On the Impact of Feature Reduction Techniques on Adversarial Attacks
特征约简技术对抗对抗性攻击的影响研究
AI 总结:本研究分析了主成分分析等特征约简技术对支持向量机等5类模型在APT数据集的3种对抗攻击下的鲁棒性影响,发现排列重要性排序和单变量互信息的鲁棒性提升最一致。
链接:https://arxiv.org/abs/2611.25004
机构:University of New Brunswick(新不伦瑞克大学)
作者:Mohammad Meymani, Roozbeh Razavi-Far
英文摘要:Adversarial attacks have posed significant threats to machine learning models by exploiting their inherent vulnerabilities. Researchers have established an interdisciplinary field called adversarial machine learning to explore these vulnerabilities and develop corresponding defense strategies. Feature reduction strategies are used to decrease the dimensionality of the data in order to reduce training time, avoid the curse of dimensionality, prevent overfitting, and mitigate redundant features. However, their effect on adversarial robustness across different classifiers and attack scenarios remains insufficiently explored. This study aims to analyze the effect of feature reduction techniques on the adversarial robustness of a diverse set of machine learning models under poisoning, backdoor, and universal adversarial perturbation attacks on the APT dataset. These models include support vector machine, decision tree, random forest, multi-layer perceptron, and k-nearest neighbors. The attained results are compared with scenarios where different feature reduction techniques are applied, including principal component analysis, permutation importance ranking, recursive feature elimination, and univariate mutual information. Metrics such as accuracy, F1-score, and mitigation score are employed to evaluate the effect of these techniques on adversarial robustness. The results demonstrate that permutation importance ranking and univariate mutual information generally provide the most consistent robustness improvements, while other methods exhibit more variable effects depending on the classifier and attack scenario.
19. TuringSpace: Grounding Spatial Intelligence in Action with Interactive CAPTCHAs
TuringSpace:通过交互式验证码将空间智能与动作建立关联
AI 总结:本研究提出TuringSpace交互式验证码环境,测试MLLMs的空间智能与动作关联能力,发现其存在误读截图、编造证据等问题,微调Qwen3.5-9B后性能提升且可迁移至真实场景。
链接:https://arxiv.org/abs/2611.25387
机构:University of Pennsylvania(宾夕法尼亚大学); University of Minnesota(明尼苏达大学); University of Southern California(南加州大学); MBZUAI(Mohamed bin Zayed 人工智能大学); University of Wisconsin–Madison(威斯康星大学麦迪逊分校)
作者:Xuwei Ding, Skylar Zhai, Linxin Song, Jiarui Feng, Dongqi Fu, Huajun Xi, Junnan Li, Jingcheng Liang, Jieyu Zhao, Yinglong Xia, Hong Li, Jian Kang
英文摘要:Humans act in the visual world by inferring the 3D layout behind a 2D view, predicting how an action changes it, and correcting themselves after seeing the result. Can multimodal large language models (MLLMs) do the same? CAPTCHAs offer a natural test, since humans solve them through this very loop, clicking and dragging on the screen and adjusting after each result. Yet existing CAPTCHA benchmarks are largely solved and rarely spatial, while spatial benchmarks rarely require action. We therefore present TuringSpace, a synthetic web environment of 125 interactive CAPTCHAs that span five spatial abilities and must be solved through clicks and drags on the screen. Every task is a program that generates new instances with a checker and an oracle, so the tasks also form a scalable training gym. Humans solve 94.1% of the tasks, whereas the best model, GPT-6 Astra, still fails 25.9% of them, and the best open-weight model, Kimi K3, fails 68.0%. Our further analysis shows that most failures come from misreading the screenshot, and that model reasoning rarely corrects a wrong belief but hallucinates evidence to defend it. To test whether such spatial abilities can be learned, we fine-tune Qwen3.5-9B on this gym. Its success rate rises from 1.6% to 33.1%, above Kimi K3, and it improves on all seven out-of-distribution benchmarks, including a 9.9-point gain on the real-world indoor videos of VSI-Bench. These results show that grounding spatial intelligence in screen actions exposes what MLLMs lack and offers a scalable way to teach it, with gains that transfer to real-world scenes. We will release TuringSpace to support further study.
20. From Sequence Uncertainty to Phenotype Hypotheses: Uncertainty-Conditioned Inference over Biomedical Knowledge Graphs
从序列不确定性到表型假设:生物医学知识图谱上的不确定性条件推理
AI 总结:本研究提出不确定性条件路径能量框架,将基因组序列预测转化为生物医学知识图谱的表型假设分布,在AlzKB等数据集上验证了其不确定性追踪、掩码边恢复及基因留存预测的优势。
链接:https://arxiv.org/abs/2611.25422
机构:New Mexico Institute of Mining and Technology(新墨西哥矿业技术学院); Cedars-Sinai Medical Center(西达赛奈医疗中心)
作者:Huixin Zhan, Jason H. Moore
英文摘要:Genomic foundation models provide variant-level evidence, but connecting that evidence to phenotype hypotheses requires accounting for uncertainty throughout structured reasoning. We introduce an uncertainty-conditioned path-energy framework that converts sequence predictions and calibrated predictive uncertainty into explicit distributions over graph-supported phenotype hypotheses, without patient-level phenotype supervision. An interpretable learned energy scores admissible gene--disease--phenotype routes, query uncertainty controls the target temperature, and path probabilities are aggregated into phenotype hypotheses. This formulation makes evidence quality an explicit determinant of hypothesis breadth while retaining traceable multi-path support. Exact normalization over enumerable candidate sets provides a reference for evaluating a Hastings-corrected learned revision operator. On AlzKB and Hetionet, phenotype-level entropy tracks upstream uncertainty, with Spearman correlations of 0.951 and 0.878, respectively; ablations identify the contribution of uncertainty conditioning. In masked-edge recovery, the model exceeds chance and degree-matched probability-permutation controls. Training the energy under the uncertainty-dependent temperature yields higher held-out recall@10 than training under a constant temperature of comparable magnitude. Under gene hold-out, recall@10 exceeds that of a trained DistMult embedding on 24 of 30 graph masks (0.183 versus 0.123); this difference is attributable mainly to three genes with many variants and is not established for all genes. These results establish an uncertainty-aware interface from genomic predictions to graph-supported phenotype hypotheses that combines interpretable probability allocation with held-out association recovery of explicitly stated scope.
9. 图学习与结构化数据 | 1 篇
21. Scone: Symmetry-aware Neural Combinatorial Optimization for Graph Coloring
Scone:面向图着色的感知对称性神经组合优化
AI 总结:该研究提出Scone模型,利用颜色等变图神经网络处理图着色问题,通过利用与打破对称性提升性能,在多任务中实现高质量解与低推理时间,且具备泛化能力。
链接:https://arxiv.org/abs/2611.25273
机构:Harvard University(哈佛大学)
作者:Paolo Pellizzoni, Melanie Weber
英文摘要:Graph coloring, which entails assigning colors to nodes such that no adjacent nodes have the same color, is a fundamental combinatorial optimization problem. In general, however, finding a node coloring with no monochromatic edges given a fixed palette is NP-hard. This motivates learning-based methods that can learn to exploit recurring structure across instances to efficiently produce high-quality colorings. Crucially, the validity of a coloring is invariant to permutations of the colors, and we show that exploiting this symmetry reduces the size of the effective state space over which the model has to learn its policy. At the same time, the coloring algorithm must be able to break some symmetries, for example to assign different colors to two adjacent nodes belonging to the same automorphism orbit. In this paper, we introduce Scone, a model that uses a color-equivariant graph neural network to construct colorings sequentially, allowing information from previous assignments to propagate and break node symmetries. Scone achieves perfect solution quality with low inference times across several graph coloring tasks, including on sudoku_extreme and register allocations, and showcases generalization to graph families not encountered during training.
10. 迁移、元学习与持续学习 | 3 篇
22. Watch Where You Linearize: Predicting and Correcting Model-Merging Interference
关注线性化位置:预测与修正模型合并干扰
AI 总结:该研究提出通过任务输出的一阶方向导数预测模型合并干扰,构建LR-Merge无标签修正方法优化合并系数,提升多任务合并性能且计算效率更高。
链接:https://arxiv.org/abs/2611.25000
机构:Tel Aviv University(特拉维夫大学); School of Electrical and Computer Engineering(电气与计算机工程学院)
作者:Ido Nitzan Hidekel, Dan Raviv
英文摘要:Model merging combines separately fine-tuned models by adding their task vectors to a shared backbone. Interference is typically assessed only after constructing and evaluating a candidate merge, making coefficient search expensive. We show that a first-order directional derivative of task A's outputs along task B's vector can predict interference, but only when evaluated at the appropriate model. At the receiver model, this derivative forecasts nearby merges per task and per example, halves the accuracy error of a no-change baseline, and identifies most examples whose predictions flip. For full multi-task merges, a single evaluation is insufficient; two evaluations along the merge path are needed. At the shared pretrained backbone, however, the derivative is nearly orthogonal to realized changes in our vision experiments and only partly predictive in LoRA language-model pilots. Controls link this failure to displacement along the receiver's own task vector, not simply parameter distance; the mismatch appears within the first few dozen fine-tuning steps. Freezing receiver-local responses produces a surrogate affine in the merge coefficients. We use it to construct LR-Merge, a label-free correction that optimizes coefficients through convex teacher matching on unlabeled probes, without backpropagation through merged networks. LR-Merge improves all tested closed-form, isotropic, and sparsified base mergers, with gains generally increasing as more tasks are merged. Using pretrained rather than receiver-local responses instead degrades the base merge. LR-Merge matches exact teacher matching through a merged network, which takes roughly three times longer to converge and performs similarly at an equal compute budget. Once built, the surrogate answers new coefficient and task-subset queries in seconds.
23. Game-Theoretic Data Allocation for Expert Specialization and Routing
用于专家专业化与路由的博弈论数据分配
AI 总结:该研究提出博弈论驱动的InflEns框架,将训练权重与推理路由统一于同一位置博弈,在安全基准上验证其能精准匹配专家与提示,攻击成功率显著降低,且无需训练显式门。
链接:https://arxiv.org/abs/2611.25167
机构:Dartmouth College(达特茅斯学院)
作者:Mark Lovett, Denis Lim, Soroush Vosoughi
英文摘要:Can the allocation used to train experts also predict which expert will perform best on a prompt? Such a connection would make the intended division of labor inspectable before training and allow the allocation itself to guide inference. We introduce InflEns, a mixture of LoRA experts whose training weights and inference routes come from the same location game in a fixed prompt-trait space. Players choose positions to maximize their expected shares of the prompt distribution, without using expert losses. We utilize the game's kernel-general theoretical criterion for when co-located positions become locally unstable to separation. We evaluate InflEns on a corpus assembled from seven safety-related benchmarks, which demonstrates that the highest-share expert has the lowest held-out response negative log-likelihood (NLL) on approximately 90% of prompts across four kernels under a sharpened simplex representation, compared with approximately 15% after permuting allocations across prompts. We demonstrate that the method performs comparably to published learned gates without having to train an explicit gate on the experts. Because the router is fixed, we can trace harmful behavior to a single expert and retrain only that expert, which lowers HarmBench attack success from 0.51 to 0.02 while keeping instruction following. The game also gives us several controls over how the experts specialize, including the kernel choice and its parameters, the trait-space geometry, and the number of experts. Importantly, InflEns is a general framework that can be extended to any encoder and data distribution.
24. Some Memories Must Fade: Persistence and Decay of User Memories in Model Parameters
有些记忆必须消退:模型参数中用户记忆的持久性与衰减
AI 总结:本研究提出一种含LoRA适配器基底与轻量级控制器的参数记忆管理系统,在含长程用户轨迹的数据集上实现86%的oracle性能,平衡召回与弃权,为长程参数记忆奠定基础。
链接:https://arxiv.org/abs/2611.25235
机构:Samsung Research America(三星美国研究院)
作者:Sanidhya Vijayvargiya, Rahul Lokesh
英文摘要:Parametric memory allows Large Language Models to internalize knowledge directly into their weights, offering a potentially powerful alternative to retrieval-based memory (RAG). However, existing approaches to parametric memory are difficult to deploy as they largely mirror the Continual Learning (CL) objective of retaining information indefinitely. This is neither sustainable nor sufficient for a memory system, which must govern knowledge across its lifecycle by retaining stable facts, reinforcing recurring observations, updating changed beliefs, and allowing transient information to fade. We propose a system for parametric memory management consisting of a memory substrate (handling information storage and forgetting) and a learned policy (what to remember, how deeply, and for how long). Our substrate is a fixed-capacity LoRA adapter that consists of k-hot rank combinations which store facts without catastrophic interference and maintain an independently controlled temporal decay, enabling outdated information to fade while long-term facts persist. On this substrate, we train a lightweight controller that governs each incoming user event to ensure the memory system reflects the user's true, updated belief state. To evaluate how memory systems manage user memory, we introduce a controlled dataset of long-horizon user trajectories with varying persistence of facts based on changing belief states and user-behavioral shifts, and probe queries that test recall while information is valid and abstention after it expires. Our policy learns to effectively manage user memory over its entire lifecycle, balancing recall and abstention to achieve 86% of oracle performance, unlike CL and RAG baselines that primarily default to one of the two behaviors. By providing the first system to manage memory directly in model weights, this work establishes a foundation for long-horizon parametric memory.
11. 数据集、基准与评测 | 3 篇
25. Long-Range Video Presentation Attack Detection: A Three-Class Benchmark Across 1-100 m and Two Sensor Types
远程视频人脸呈现攻击检测:1-100米及两种传感器类型下的三分类基准
AI 总结:该研究针对远程人脸PAD场景,构建含261个身份的视频数据集,测试四种CNN模型,发现EfficientNet性能最优,验证了跨域鲁棒性,为远程PAD研究提供基准。
链接:https://arxiv.org/abs/2611.25108
机构:Clarkson University(克拉克森大学); UNC at Charlotte(北卡罗来纳大学夏洛特分校); Ombra LLC(奥姆布拉有限责任公司)
作者:Lambert Igene, Mohammad Zahir Uddin Chowdhury, Afzal Hossain, Sandip Purnapatra, Peter Johnson, Stephanie Schuckers
英文摘要:Face presentation attack detection (PAD) systems deployed in border control and airport surveillance are constrained by a critical assumption: subjects are captured at close range, typically under one meter. No existing benchmark systematically addresses multi-distance capture, diverse environmental conditions, and concurrent spoofing and occlusion-based attacks, leaving deployed models underprepared for real-world scenarios. We introduce a video-based PAD dataset of 374,000+ frames from 261 unique identities collected across five standoff distances (1 m-100 m) under indoor and outdoor conditions using a DSLR camera (C) and a mobile sensor (M), with three-class labels: live, spoof, and props (occlusions); models are trained and evaluated on video clips rather than single frames. We fine-tune and benchmark four CNN architectures: MobileNetV3, MobileNet, EfficientNet, and ResNet-50. The higher-resolution DSLR maintains near-perfect spoof classification rate across all distances, while the lower-resolution mobile sensor degrades sharply beyond 50 m, with props classification rate dropping to 55% at 75 m for MobileNetV3. EfficientNet achieves the best overall performance (98.48% accuracy, ACER 2.01%). Generalization tests on unseen attack datasets yield 89-94% accuracy, validating cross-domain robustness. These results establish capture distance and sensor resolution as critical but underexplored axes of PAD difficulty. The dataset will be publicly released under an End-User License Agreement to support reproducible long-range PAD research.
26. Moving Generality from Model to Modeler: LLM Agents for PHM on a Deployment-Style Benchmark
将通用性从模型转移到建模者:面向部署式基准的PHM大语言模型智能体
AI 总结:该研究在PHM领域提出,通用性从模型转移到建模者,NextUnit-Bench基准实验显示前沿LLM智能体优于AutoML及部分已发表方法,其优势源于数据驱动的设计与学习能力。
链接:https://arxiv.org/abs/2611.25252
机构:CentraleSupélec, Université Paris-Saclay(中央高等电力学院 巴黎萨克雷大学)
作者:Zhenling Chen, Zhiguo Zeng, Anne Barros
英文摘要:Coding agents built on large language models (LLMs) now help develop many models published as general solutions. Should generality instead reside in the agent that builds a model for each new task? We examine this question in prognostics and health management (PHM), where every model must work on unseen equipment. Our deployment-style benchmark, NextUnit-Bench, covers 24 public datasets split into 58 folds that hold out units or operating conditions; each agent has two hours to deliver an executable predictor, then run on the hidden units. First, frontier agents outperform AutoML, which reaches less than half the best agent's score, and faithful implementations of published methods on four datasets, of which only 6 of 30 beat the strongest agent's own design; published accuracies, where reproducible, rest on the papers' own conventions, and a single convention, capping the remaining life of test engines, can halve the reported error. Second, the agents' advantage is a design made on each machine's data; once made, a plain-text plan of a few hundred words carries most of the gap between two frontier agents, and given a weaker agent's plan, the strongest agent falls to that agent's level, often to identical predictions. Before touching the data, however, it matters little whose plan an agent carries out, and domain toolkits appear to help neither: what sets the strongest apart is how much it learns from the data. Third, such designs reach only implementers that can execute them: given the strongest agent's designs, a locally deployable open-weight model delivers half as often, even when they are lightened, and scores no higher than without them. Generality, we argue, is not disappearing but moving, from the model to the modeler.
27. FG-GraphBench: Benchmarking and Enhancing Fine-Grained Visual Understanding with Multimodal Scene Graphs
FG-GraphBench:基于多模态场景图的细粒度视觉理解基准构建与增强
AI 总结:该研究针对现有场景图基准覆盖不足及MLLMs未有效利用图结构信息的问题,提出FG-GraphBench基准与FiGRA方法,提升了细粒度视觉理解性能。
链接:https://arxiv.org/abs/2611.25408
机构:Beijing Institute of Technology(北京理工大学); Shandong University(山东大学)
作者:Sicheng Liu, Xunkai Li, Ziyu Han, Xingnan Zhang, Ru Zhang, Rong-Hua Li
英文摘要:Fine-grained visual understanding requires multimodal large language models (MLLMs) to recognize small visual entities and understand their relationships. Scene graphs offer a natural way to represent this information, organizing visual entities as nodes and their relationships as edges. However, we identify two challenges: existing scene graph benchmarks provide insufficient coverage of fine-grained entities and the relationships among them, and existing methods for fine-grained visual understanding have yet to effectively leverage such graph-structured information in MLLMs. To address these challenges, we introduce Fine-Grained Multimodal Scene Graph Benchmark (FG-GraphBench) and FiGRA. FG-GraphBench is a benchmark comprising three complementary tasks: fine-grained visual question answering (FG-QA), image captioning (FG-Caption), and image description generation (FG-Description). We develop a two-stage construction pipeline that discovers and grounds fine-grained entities and establishes their relations to construct the fine-grained multimodal scene graphs (MSGs) in FG-GraphBench. Building on these MSGs, we propose FiGRA, a Fine-grained Graph Representation Augmentation method for MLLMs. FiGRA adopts a two-stage training strategy. It encodes multimodal entity and relation features as continuous graph tokens, and integrates them into the MLLM. Experiments show that FiGRA significantly outperforms strong baselines across three tasks of FG-GraphBench. Further analyses confirm the effectiveness of FG-GraphBench and highlight the importance of preserving fine-grained entity and relation information for visual understanding.
12. 机器学习应用 | 4 篇
28. Auditing Learned Dynamics in Weather Models
对天气模型中学习到的动力学进行审计
AI 总结:本研究以GraphCast为对象,检验天气模型的因果相关性与物理语义的关系,通过操控稀疏自动编码器特征,发现其可影响极端事件预报误差,为审计学习到的天气动力学提供了实验坐标。
链接:https://arxiv.org/abs/2611.25119
机构:Mila(米拉研究所); McGill University(麦吉尔大学); Haziel(哈齐尔公司)
作者:Skylar Gu, Christina Last
英文摘要:Learned weather models are experimentally accessible computational systems whose internal representations can be inspected and manipulated. Using sparse autoencoder features from GraphCast, DeepMind's autoregressive weather forecasting model, we test whether causal relevance requires physical semantics, a question that matters for auditing whether AI forecasters used in high-assurance settings rely on reliable mechanisms. We also look at whether physically associated latent directions can be steered to improve extreme-event forecasts, where interventions are applied to the model's internal activations during the rollout. We find that ablating a feature locked to GraphCast's computational mesh increases global geopotential height error despite lacking conventional physical semantics. Second, suppressing convection- and low-level-vorticity-associated features weakens predicted tropical-cyclone deepening (the fall in central pressure during intensification), while amplifying them reduces intensity error, with amplification of convection features through feature scaling achieving a median of 51\% error reduction over seven storms. Together, these results separate interpretable, physical association from causal utility and establish sparse features as experimental coordinates for auditing and intervening on learned weather dynamics.
29. Temporal Extrapolation of Elastic Seismic Wave Propagation in Layered Media Using Neural Operators
基于神经算子的层状介质弹性地震波传播时间外推
AI 总结:本研究提出耦合神经算子框架(含TI-DeepONet与FNO),实现层状弹性介质地震波时间外推,在4个算例中预测误差低,验证了该方法的可行性。
链接:https://arxiv.org/abs/2611.25130
机构:Johns Hopkins University(约翰斯·霍普金斯大学); University of Utah(犹他大学)
作者:Dibyajyoti Nayak, Yuze Pu, Kami Mohammadi, Somdatta Goswami
英文摘要:Temporal prediction of seismic velocity fields beyond the time interval represented in training remains challenging for neural operator models. This study develops a coupled neural operator framework for stress-conditioned temporal prediction in layered elastic media. A time-integrator-embedded Deep Operator Network (TI-DeepONet) advances the velocity field through an embedded fourth-order Runge-Kutta (RK4) scheme, while a Fourier Neural Operator (FNO) learns a residual correction to the predicted spatial field. The two components are trained jointly in a predictor-corrector configuration. Prediction is initialized with reference velocity fields at the first two time levels and proceeds autoregressively, with reference stress fields and prescribed excitation supplied throughout rollout, including beyond the training interval. Four cases consider horizontally polarized shear (SH) waves in four-layer soil columns with fixed total depths of 50 and 280 m, excited by third-order Gaussian-derivative (DOG3) wavelets with central frequencies between 0.5 and 6 Hz. Layer thicknesses and shear-wave velocities vary across profiles. Evaluations on test profiles extend beyond the training interval by up to twice its duration. The predictions reproduce principal velocity arrivals and depth-dependent response patterns, with accuracy varying across profiles. Across the four cases, mean peak-normalized root-mean-square errors (NRMSEs) range from 0.53% to 2.31% over the full evaluation windows and from 0.61% to 2.82% over the extrapolation intervals. Both metrics use each profile's peak absolute reference velocity over the full evaluation window as the normalizer. These results demonstrate temporal velocity prediction under reference-stress conditioning for the tested configurations; autonomous continuation of the coupled velocity-stress wavefield remains to be established.
30. PDE-State Conditioning for Learned Optimization Using Physics-Informed Neural Networks
基于物理信息神经网络的学习型优化中的PDE状态条件设置
AI 总结:本文针对学习型PINN优化器未利用PDE信息的问题,提出基于PDE状态的条件设置方法,构建11维PDE状态结构,采用残差几何状态G2和有界FiLM,使优化器在多个PDE问题上的评估指标优于Adam和无条件学习型优化器。
链接:https://arxiv.org/abs/2611.25185
机构:University of Houston(休斯顿大学); University of Austin(奥斯汀大学)
作者:Farzana Yasmin, Ricardo Vilalta
英文摘要:Physics-informed neural network (PINN) optimization depends on global properties of the underlying PDE and how the physics residual evolves across the domain. Standard optimizers such as Adam use fixed gradient-based statistics, while learned PINN optimizers replace the fixed rule with a learned update built from local parameters and optimization-history features. However, none explicitly use PDE-level information. We study two coupled design questions: what PDE-level information should a learned optimizer observe, and how should that information influence its update? We introduce an 11-dimensional PDE state-capturing operator structure, residual geometry, and short-horizon update response, and use it to condition a structured meta-learned optimizer through bounded feature-wise modulation. Repeated design studies identify the compact residual-geometry state G2 and bounded FiLM as the strongest overall choices among those tested. The resulting PDE-State-conditioned optimizer improves all evaluation metrics (relative-L2, L1, solution MSE, and physics MSE) over Adam and the unconditioned learned optimizer on Linear Advection, KdV, and Burgers.
31. Intelligent Embedding Pruning for Large-Scale Recommendation Models
面向大规模推荐模型的智能嵌入剪枝
AI 总结: 该研究针对大规模推荐系统嵌入层内存压力问题,提出基于特征重要性的训练中嵌入剪枝方法FI-ITEP,可实现20%-40%内存减少,提升GPU利用率并降低训练成本。
链接:https://arxiv.org/abs/2611.25288
机构:Meta
作者:Anish Khazane, Alex Saad-Falcon, Guanchu Wu, Marcio Porto, Jasper Shan, Fei Yu, Eugene Shulga, Intaik Park, Yongzhe Zhang, Xiaoyu Deng, Sheng Luo, Xiaoya Wang
英文摘要:Modern recommendation systems face substantial memory pressure, largely from the sparse embedding layer that maps high-cardinality categorical features into dense vectors. In-training embedding pruning (ITEP) removes less important embedding rows during training, yielding end-to-end memory savings but sometimes degrading accuracy because per-table pruning ratios are set by size-based heuristics. We use feature importance (FI) to guide these ratios (FI-ITEP), assigning higher pruning to tables tied to less important features. Across three large-scale recommendation models, FI-ITEP achieves 20--40% total memory reduction while matching or improving accuracy relative to percentile-based ITEP. Because the feature-importance signal requires minimal resources to compute, these memory savings translate into concrete systems benefits: the same models train on fewer GPU trainers (e.g., 112 to 80) with higher per-GPU throughput, saving 1,000+ GPU-hours per model launch. We reproduce these trends on the open-source DLRM (Criteo 1TB) and quantify the end-to-end training-time and GPU-cost tradeoff---including the burden of computing FI---offering practical guidance for embedding pruning in real-world systems.
13. 其他/综合机器学习 | 9 篇
32. PsyWorld: Profile-Conditioned World Modeling for Human Behavior Prediction
PsyWorld:用于人类行为预测的画像条件化世界建模
AI 总结:本研究提出PsyWorld画像条件化世界模型,利用环境-行为状态转移学习表征,在VEET和AgentSociety中评估,实现对未见个体的泛化行为预测,取得良好性能。
链接:https://arxiv.org/abs/2611.24999
机构:Emory University(埃默里大学)
作者:Ziyuan Qin, Yang Qiao, Summer B. Frandsen, Michael T. Treadway, Liang Zhao
英文摘要:Understanding how stable psychometric traits and time-varying states jointly inform behavior is an important problem in computational behavioral science. Existing personalized agents commonly condition actions directly on a static profile, leaving realized state transitions unused as an explicit source of supervision. We introduce a profile-conditioned state world model that uses observed environment--behavior state transitions to learn transition-informed representations for behavioral prediction. During training, a posterior encoder quantizes realized transitions, while a prior predicts the resulting codes from the psychometric profile and pre-transition context. The prior code and predicted next state then condition macro-level behavior and simulator-grounded execution. Thus, the world model provides structured transition supervision instead of treating personality as a direct action descriptor. We evaluate the framework in VEET, a controlled behavioral simulator with human trajectories, and in AgentSociety as a structurally different synthetic environment. A participant-disjoint split of 240 training, 25 validation, and 48 test participants evaluates generalization to unseen people. On VEET, the model reaches 0.904 Top-1 accuracy (95\% CI: [0.895, 0.913]) and 0.0033 Macro KL. Ablations associate continuous next-state modeling with most of the Top-1 gain, while discrete codes and prior alignment mainly improve Top-3 and distributional alignment. These results establish transition-informed world-model supervision as useful for subject-generalized behavior prediction, without implying causal psychological mechanisms or calibrated long-horizon state dynamics.
33. A Model of Diverse Sampling from Language Models
语言模型的多样化采样模型
AI 总结:针对语言模型生成样本集合质量差的问题,提出PoolDPP方法,通过结合行列式点过程(DPP)与候选池采样,在多个任务上拓展了质量-多样性等性能边界并提升了多样性。
链接:https://arxiv.org/abs/2611.25003
机构:ELLIS Institute Tübingen(埃利斯研究所蒂宾根分支); Mila(米拉研究所); ETH Zürich(苏黎世联邦理工学院); Cornell University(康奈尔大学); McGill University(麦吉尔大学)
作者:Manuel Prada-Corral, Yahya Emara, Timothy J. O'Donnell, Ryan Cotterell, Tim Vieira
英文摘要:Language models often produce high-quality individual samples but poor sets, with repeated samples clustering around the same semantic modes. We formalize diverse generation as sampling a size-$K$ subset of complete strings from a determinantal point process (DPP), where the likelihood encodes item quality, and the embedding geometry encodes repulsion between similar outputs. Exact inference over all strings is intractable, so we propose PoolDPP: draw a finite candidate pool from a tractable proposal, importance-weight the candidates toward the target quality, and run a $K$-DPP sampler on the induced pool kernel; a greedy MAP variant provides deterministic selection. We prove a finite-sample bound with an explicit constant: the KL divergence between the sampler's output and the target $K$-DPP decays as O(1/N) in the pool size. Empirically, PoolDPP extends the quality--diversity frontier on open-ended generation and the precision--recall frontier on ambiguous text-to-SQL (Ambrosia), and on mathematical reasoning (GSM8K), it raises diversity at about the same pass@$K$.
34. Fast and Memory-Light Evaluation of Density-Based Clusterings with the DISCO Index in R
在R中使用DISCO指数快速且低内存地评估基于密度的聚类
AI 总结:该文介绍R包discoCVI的设计,其实现的DISCO指数可低内存、快速评估基于密度的聚类,用于无真实标签时的聚类比较与HDBSCAN参数选择,性能优于旧版本及Python参考实现。
链接:https://arxiv.org/abs/2611.25114
机构:Università di Milano-Bicocca(米兰比可卡大学)
作者: Amin Entezari
英文摘要:DISCO is a recent internal cluster validity index for density-based clustering that, unlike earlier indices, also evaluates the quality of noise labels. The discoCVI package makes DISCO available in R. This article describes the design of the package and a typical workflow, in which DISCO is used to compare clusterings and to choose the parameters of HDBSCAN without ground truth, followed by the implementation introduced in version 0.2.0. DISCO is defined through density-connectivity distances between all pairs of points, so direct implementations store several n-by-n matrices. We show that every quantity DISCO needs can be read off the single-linkage merge tree of the mutual-reachability minimum spanning tree, so that the index can be computed exactly with memory that grows linearly with the number of points, while the running time remains quadratic as in the original definition. The implementation is tested against the previous version, agrees with the original Python implementation to within $3 \times 10^{-13}$, and corrects the score of singleton clusters to match the published definition. In benchmarks, discoCVI is 23 to 138 times faster than its previous version for 500 to 4,000 points and 3 to 9 times faster than the Python reference for 500 to 8,000 points, using a fraction of the memory of both, and it scores 100,000 points in about 50 seconds with less than 200 MB of memory.
35. Rate-distortion Assessment for Hierarchical Content Summarization
分层内容摘要的率-失真评估
AI 总结:该研究受率-失真理论启发,将分层内容摘要形式化为信息编码问题,设计框架提取源内容语义层次结构以量化率-失真权衡,通过实验验证其在大语言模型内容摘要及检索增强生成中的有效性。
链接:https://arxiv.org/abs/2611.25151
机构:SRI International(SRI国际)
作者:Yunye Gong, Indranil Sur, Jared Claypoole, Pritish Sahu, Ajay Divakaran
英文摘要:Powerful generative models including large language and vision models have demonstrated success across a wide range of content generation, editing and manipulation tasks via end-to-end prompt-based generation. However, current assessment of the quality of generative content manipulation remains subjective, limiting scientific understanding of model capabilities. Inspired by rate-distortion theory from Shannon information compression theory, in this work we formalize hierarchical content summarization as an information encoding problem. We design a principled framework to extract the semantic hierarchy of the source content to support quantitative characterization of the rate-distortion trade-off between compression efficiency (rate) against information loss (distortion) during generative content manipulation, which allows us to benchmark various models with respect to performance conditional on rate budget of interest. We demonstrate the use of the framework with empirical experiments on content summarization for long-context question-answering with state-of-the-art large language models, offering new insights into their efficiency and comprehension capabilities. We further demonstrate the use of the semantic hierarchy as guidance for performing effective retrieval augmented generation, showing improved rate-distortion characteristics against current baselines.
36. Learning to Read Events for Exact State Tracking with Frozen Language Models
利用冻结语言模型学习读取事件以实现精确状态跟踪
AI 总结:本研究在冻结语言模型上训练选择器读取事件,结合外部整数寄存器实现精确状态跟踪,在多任务测试中较基线提升性能并降低时间与成本。
链接:https://arxiv.org/abs/2611.25211
机构:Attractor Dynamics Inc.(吸引子动力公司)
作者:Jeonghoon Lee (Attractor Dynamics Inc.)
英文摘要:Long event streams require identifying each change and applying the changes in order. We train a selector on a frozen language model to read each transition. With API-only access, we ask the model to list the transitions. An external integer register composes them in order. A confidence rule can flag uncertain updates or retain several candidates. We study products of permutations of five items, a task that fixed-depth, log-precision transformers cannot solve on all input lengths under the standard conjecture that $\mathrm{TC}^0 \ne \mathrm{NC}^1$. On two-event windows, a 27B model answered 3 of 100 correctly without visible reasoning and 96 of 100 when it generated intermediate steps. A frozen 0.8B model with a trained selector read every event in 299 of 299 windows of 43,687 events each. The same package passed 298 of 299 new windows on a fresh machine, missing the target of 0.99 for the one-sided 95% lower confidence bound. A separate follow-up sized to allow one failure yielded 472 of 473. On 1,000-event room-swap documents, three of ten models solved all five documents by reasoning alone. All seven models tested both ways at this length solved all five with extraction and the register, with no chain of thought requested. With concurrent extraction, median wall-clock time fell from 136.1 to 10.4 s for Sonnet 5.5 and from 313.5 to 28.0 s for GPT-6.1 Sol. At fixed per-token rates, mean cost per document fell by 55% for Sonnet 5.5 and 44% for GPT-6.1 Sol. A regular expression with an executor also solved all 15 fixed-format documents. On 500 free-form issue-tracker comments, a separately adapted 4B reader identified the recorded value in 68.8% of cases, compared with 61.0% for a frequency baseline fitted on the same training projects.
37. Prescribed Bayesian Efficiency under Lifetime Adjustment Costs
寿命调整成本下的规定贝叶斯效率
AI 总结:该研究在寿命调整成本下,刻画了贝叶斯学习器累积预期超额与有限预期寿命费用的联合可达性,提出确定性检查点策略等方法,验证了逻辑决策模型的假设。
链接:https://arxiv.org/abs/2611.25311
机构:Institute for Artificial Intelligence(人工智能学院); VNU-University of Engineering and Technology(越南国家大学工程技术大学)
作者:Viet Bac Nguyen
英文摘要:
A Bayesian learner may retain an issued action while updating its posterior, trading predictive loss against adjustment charges. We assume population curvature, boundary control, an order-$1/n$ lower bound on expected posterior uncertainty about the optimal action, and an observable unbiased gradient witness whose squared norms are uniformly integrable. Sample losses may be unbounded or nonconvex. Cumulative expected excess measures the loss increase relative to current predictive Bayes actions. For bounded nonnegative nondecreasing charges $c$ with $c(0)=0$ and nondecreasing profiles $G: \mathbb{N} \to [1,\infty)$ with $G(J)=o(J)$, we characterize joint attainability of cumulative expected excess $O(G(J))$ at horizons $T=2^J$ and finite expected lifetime charge. Here $J$ indexes logarithmic time. The criterion is finiteness of a prefix-constrained allocation of tracking-error allowances across action-distance scales, equivalently its decreasing-weight dual. Adaptive timing and private randomization do not enlarge the attainable region. For each feasible pair, a horizon-independent deterministic checkpoint policy attains the profile with lifetime-cost moments of every positive order. For excess $O((\log T)^r)$, $0
38. Now You See Me Now You Don't: A Framework To Detect and Reduce Proxy Capacity
《现在你看到我,现在你又看不到:一种检测并降低代理能力的框架》
AI 总结:该研究针对受监管领域的差异性对待问题,提出两部分组成的框架,可检测并降低特征的代理能力,同时保留预测性能,优于直接移除代理特征的方法。
链接:https://arxiv.org/abs/2611.25327
机构:Upstart Network, Inc.(Upstart网络公司)
作者:Nirek Sharma
英文摘要:In regulated domains such as lending, the two most salient fairness concepts are disparate impact and disparate treatment: disparate impact is generally defined by differential rates of adverse outcomes between demographic groups, while disparate treatment is generally defined by the use of demographic information or proxies to issue decisions. There already exist established methods to quantitatively detect and mitigate disparate impact; however, there has been less research on doing the same for disparate treatment. From a practitioner's perspective, the manner in which a conditional independence test is formulated and operationalized to identify proxies is nuanced and underexplored. As Machine Learning (ML) systems increase in scope and application in regulated domains, practitioners need a defensible framework to identify proxies at scale. Current disparate treatment frameworks typically rely upon some combination of quantitative and qualitative review of models' inputs to identify proxies. Features that are identified as proxies are then typically removed from the model in practice, and predictive performance degrades in favor of fairness through unawareness. Here, we propose a two-part framework that reduces the disparate treatment risk of a feature and keeps the predictive performance that removal of the feature loses. First, we propose a quantitative test for proxy capacity that identifies whether some of a feature's values reveal membership in a particular demographic group. Second, we propose a pre-processing methodology that adjusts those proxy features to reduce their proxy capacity. We show empirically on synthetic and public datasets that this framework identifies features that proxy for demographic information and reduces their proxy capacity while the adjusted feature retains task-relevant signal (i.e. improves predictive performance versus removal).
39. Class-Conditional Conformal Triage for ICAO Face-Image Compliance under Class Imbalance
类别条件共形分诊用于ICAO人脸图像合规性(类别不平衡场景下)
AI 总结:本研究针对类别不平衡下的ICAO人脸图像合规性,将共形预测作为BioGaze的分诊层,发现类别条件共形预测可提升少数违规类别的覆盖率,同时揭示了统计覆盖率、分类性能与重拍需求的权衡。
链接:https://arxiv.org/abs/2611.25342
作者:Mustafa Vedat Yildirim, Inna Skarga-Bandurova
英文摘要:Automated ICAO/ISO face-image compliance analysers provide criterion-level pass/fail decisions, yet these verdicts do not quantify uncertainty when deciding whether to accept, reject, or retake a photograph. We investigate conformal prediction as a triage layer for BioGaze, evaluating 14 compliance criteria on the synthetic TONO benchmark. Without retraining the analyser, we fit per-criterion logistic classifiers and compare marginal and label-conditional (Mondrian) conformal prediction with conventional abstention rules using subject-disjoint splits. Marginal conformal prediction achieved approximately 0.90 overall coverage but substantially under-covered violations, with a ten-partition mean of 0.37 for one criterion. The corresponding Mondrian means ranged from 0.89 to 0.92 across criteria. However, most of the reduction in criterion-level false positives came from the logistic layer (0.145 to 0.048), while tuned abstention rules achieved lower false-positive rates than Mondrian at comparable retake rates. Bonferroni correction for joint coverage across 14 criteria required retaking nearly all images. These findings show that class-conditional conformal prediction can achieve near-target empirical coverage for minority violation classes and reveal trade-offs between statistical coverage, classification performance, and retake requirements in multi-criterion face-image compliance assessment.
40. Antom 3-in-1 Transformer: A Payment Foundation Model
Antom三合一Transformer:支付基础模型
AI 总结:本研究开发了Antom三合一Transformer支付基础模型,其整合多类表示,经预训练后在跨业务迁移、少标签场景及在线实验中均展现出优异的支付风险检测性能。
链接:https://arxiv.org/abs/2611.25403
机构:Ant International(蚂蚁国际); East China Normal University(华东师范大学); Shanghai Jiao Tong University(上海交通大学); Tongji University(同济大学)
作者:Weiyu Han, Haoyu Jiang, Jie Sheng, Youxi Pan, Xianglong Liu, Zeqi Ren, Dongquan Yi, Yinhao Liu, Ruobing Zuo, Ze Zhao, Huilin He, Nairong Zhang, Xinyi He, Xiao Tan, Jun Wang, Pan Fang, Wei Zeng, Jianan Su, Kaixin Shen, Jingtong Wu, Hao Li, Shuai Chen, Jiangming Yang, Tianyi Zhang
英文摘要: We report the development of Antom 3-in-1 Transformer, a payment foundation model that integrates sequence, tabular, and graph representations for payment risk detection. The model contains approximately 7B parameters, with approximately 1.7B activated per task route. Antom 3-in-1 Transformer is pretrained to learn reusable behavioral and relational representations, primarily on Ant International's internal payment data. In a cross-business transfer experiment, pretraining improves the target business metric by 34.5% over the same architecture trained from scratch. With only 25% of the training labels, the pretrained model outperforms the full-label scratch baseline in Average Precision (AP) and Recall@0.5% on different merchant domains. In a four-month online experiment on Ant International's billion-scale payment platform, the model reduced the overall loss rate by 37.88% relative to XGB-Online.
因公众号篇幅限制,更多论文请扫码查看