跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

覆盖计算机全学科，以 AI 为快捷入口。关注方向、作者和关键词，把发现变成持续的研究积累。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1072新增论文
+845其中 AI 相关
+94已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-09 16:31 · 全站最近入库：2026-11-09 13:22各信源进度 →

统计口径与数据新鲜度

统计 2026-11-08 16:41 — 2026-11-09 16:41（北京时间） · 点击数字查看对应简报；AI 相关包含在新增论文内。

最近完成检查：2026-11-09 16:31最近发现新增版本：2026-11-09 13:22库内论文最新更新时间：2026-11-07 02:59

每分钟刷新已保存的数据。空检查不代表采集异常；新收录可能包含历史论文。AI 导读按新增版本计数。

18,172 已收录论文15,506 AI 相关4,527 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

时间范围跟随排序：版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常，查看状态 →

找到 18,172 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-09 16:29（北京时间）
新论文 · v1arXiv:2610.37443

ETAMem: Decoupling and Aligning Heterogeneous Memory for Long-Term Conversational Agents

Xinying Qian, Ying Zhang, Yu Zhao, Xubo Liu, Baohang Zhou, Jeff Z. Pan

原文摘要

Long-term memory enables Large Language Models (LLMs) to maintain coherent and personalized interactions across extended conversations. However, existing memory systems face two key challenges: (1) Information Entanglement, where heterogeneous information is mixed within coarse-grained memory units; and (2) Organizational Mismatch, where memory structures based primarily on semantic similarity may overlook other associations. To address these challenges, we propose ETAMem, a multi-view memory framework that decouples heterogeneous conversational information and aligns memory representations and organization with its distinct characteristics. Specifically, heterogeneous information parsing extracts atomic memory units and identifies affective information. View-specific memory structuring then organizes these units into three complementary hierarchical views: Event, Affect, and Time. ETAMem incrementally maintains memory structures with adaptively calibrated thresholds. For retrieval, entropy-based retrieval allocates query-specific budgets across views and performs hierarchical search within each view. Experiments on LoCoMo and LongMemEval-S demonstrate that ETAMem outperforms state-of-the-art baselines, achieving relative accuracy improvements of 12.46% and 7.96%, respectively, while maintaining competitive token efficiency. Further analyses demonstrate the complementary benefits of different memory views.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37442

CRISP: Co-designing Pruning and Scheduling for Efficient MoE Inference on Edge Servers

Congwei Zhang, Youming Tao, Yifei Zou, Yu Yang, Ruirui Zhang, Xiaoyu Zhang, Falko Dressler, Xiuzhen Cheng 等 9 位作者

原文摘要

Mixture-of-Experts (MoE) inference on memory-constrained edge platforms is often dominated by expert-weight movement rather than computation. Although each token activates only a small subset of experts, routing is input-dependent, so all experts must remain accessible and memory movement dominates latency. Existing approaches such as uniform pruning and expert dropping overlook two properties of MoE execution: expert combinations are highly skewed in frequency, and experts contribute asymmetrically within each combination. We present CRISP, a combination-aware framework that co-designs pruning and runtime scheduling for efficient MoE serving on memory-constrained edge platforms. CRISP reorders neurons by importance so one stored expert can support multiple execution widths, assigns differentiated execution to hot, warm, and cold combinations based on frequency and expert role, and exploits heterogeneous widths to overlap expert loading with computation. Across Mixtral-8x7B and DeepSeek-MoE-16B, CRISP improves throughput by up to $2.1\times$ at the same memory budget while limiting accuracy loss to 3.6\% relative to full-width inference. Our results highlight that efficient edge MoE serving requires jointly optimizing which expert capacity is preserved and how expert execution is scheduled.

cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.05758

DRIFT: A Residual Flow Adapter for Decoding Continuous Outputs in Vision-Language Models

Zhuoming Liu, Jinhong Lin, Kwan Man Cheng, Lin Zhang, Shayok Bagchi, Yin Li

原文摘要

Many modern vision-language models (VLMs) build on autoregressive decoding of discrete tokens. While text-based output interfaces enable scalable pretraining and strong zero-shot generalization across diverse tasks, they are poorly suited for problems that require precise continuous outputs, such as localizing temporal boundaries of events or generating robotic control actions. To address this challenge, we propose DRIFT, a general framework for adapting pretrained VLMs to continuous decoding tasks. DRIFT combines a base predictor, which provides a coarse estimate of the target output, with a generative refinement module based on flow matching that iteratively improves the prediction. This residual formulation transforms the generative modeling problem from learning a global output distribution to modeling a localized residual distribution around a strong prior, substantially simplifying optimization. We evaluate DRIFT on both perception and planning tasks, including visual grounding and robotic control. Across multiple tasks and architectures spanning MLLMs, VLAs, and WAMs, DRIFT consistently outperforms a strong set of regression- and generative-based solutions.

cs.CV · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.34593

What Does World Modeling Actually Improve in Robot Action Learning?

Hanwen Cui, Yuanyuan Li, Qing Guo, Zhida Zhang, Chao Chen, Xi Chen, Jie Cao

原文摘要

World Action Models (WAMs) combine future-video prediction with robot action learning, yet improvements in task success alone do not reveal how predictive supervision manifests in execution behavior or whether its benefits persist under task and environmental shifts. We conduct controlled future-video prediction loss ablations within FastWAM and OpenWAM, keeping model architecture, initialization, robot training data, and training steps fixed within each model family. Under clean conditions on seen tasks, predictive supervision increases success rates by 24.38 and 72.91 percentage points, respectively. Trajectory clustering further reveals higher success rates within several coarse trajectory clusters, including cases where cluster frequencies change little, indicating that the gains are not solely reflected in changes in the prevalence of trajectory patterns. However, these benefits do not consistently persist under distribution shifts: improvements on unseen tasks vary across models and tasks, while both models exhibit low success rates under the evaluated environmental randomizations, without a consistent advantage from predictive supervision. We further introduce future-video prediction as an auxiliary training objective for $π_{0.5}$, where a lightweight Transformer predictor used only during training increases success from 82.94% to 89.19%. These findings support future-video prediction as effective auxiliary supervision for action learning, while distinguishing improved execution on familiar tasks from reliable transfer and showing that such gains need not rely on a large pretrained video diffusion transformer (DiT) in the evaluated setting.

cs.RO

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37441

Multi-Agent Reinforcement Learning of Co-Adaptive Hip Exoskeleton Control (MARL-Exo) for Metabolic Energy Reduction

Yifei Yuan, Jakob Wolf, Ghaith Androwis, Xianlian Zhou

原文摘要

Objective: This study aims to determine whether a hip-exoskeleton policy trained in simulation, by co-adapting with a musculoskeletal human model under a metabolic-energy reward, reduces the metabolic cost of walking in new users without per-user tuning, and how mechanical-power shaping affects the learned assistance. Methods: We developed MARL-Exo, a multi-agent reinforcement learning framework in which a full-body human model (304 muscle actuators; randomized anatomy, gait, and speed) and a bilateral hip-exoskeleton policy are trained concurrently in two stages. The exoskeleton observes only delayed hip kinematics and its past commands; its reward includes muscle metabolic power from a physiologically based energetics model, alone (Metabolic) or with mechanical-power shaping (Metabolic + Power). The Metabolic policy was deployed and tested without retraining in eight healthy adults walking at 1.25 m/s, using indirect calorimetry. Results: In simulation, the Metabolic policy reduced modeled muscle metabolic power by 13.8% relative to no-device walking, versus 8.8% for Metabolic + Power. On hardware, MARL-Exo reduced net metabolic rate by 18.3 $\pm$ 6.6% relative to Passive in all eight participants. Relative to No-Exo, the mean reduction was 7.4 $\pm$ 10.0%; six participants more than offset the 13.1% device-wearing penalty, walking up to 24.2% below their No-Exo cost. Conclusion: A metabolically informed policy co-adapted in simulation transferred to users without per-user tuning and consistently reduced the metabolic cost of walking with the device, while explicit power shaping reduced the simulated benefit. Significance: Unlike per-user human-in-the-loop optimization, this approach yields a single population-level controller and allows assistance objectives to be screened in simulation before human testing.

cs.RO · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37437

Representing the Unseen: Rethinking Visual Imagination for Spatial Reasoning

Zengqi Zhao, Zun Wang, Shoubin Yu, Mohit Bansal, Yue Zhang

原文摘要

Visual world models increasingly support spatial reasoning by generating RGB views of unobserved states. However, RGB has become the default interface largely because it is well suited for visual generation, rather than necessarily being the most effective representation for downstream reasoning. Different spatial questions may benefit from different representations of the same imagined state: top-down geometry can facilitate global distance comparisons, while panoramic views provide broader context for reasoning about open space. To isolate the role of representation in spatial imagination, we introduce ImagiRep, a simulator-grounded benchmark of 922 questions spanning seven question types across two settings, Viewpoint Imagination and Cross-view Integration, providing oracle RGB and six alternative representations (e.g., depth, panoramic views, and top-down geometry) of the same unobserved state. This controlled comparison reveals that even accurate imagined RGB does not guarantee accurate spatial reasoning, and different questions benefit from different representations. Similar trends hold for representations derived from world-model-estimated states. Notably, estimated RGB alone can mislead spatial reasoning, while alternative representations retain complementary benefits. Building on these findings, we introduce GEARS, a lightweight test-time strategy for adaptive representation selection, outperforming existing spatial imagination methods and showing that reasoning about unseen states requires more than generating RGB views.

cs.CV · cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.24045

What Remains Without Leaking Test Labels? Exposing Evaluation Flaws in Graph-Level Anomaly Detection

Yujie Lu, Da Yan

原文摘要

Graph-level anomaly detection (GLAD) identifies unusual graphs in applications such as molecular screening, often using only normal graphs for training. Yet the evaluations of some detectors select models with test labels, and high benchmark scores alone do not show whether detectors recognize structural changes. We audit released code and compare the area under the receiver operating characteristic curve (AUROC) for model states (checkpoints) chosen with and without test labels from the same training run. We separately compare test-tuned hyperparameters with repository defaults and test sensitivity to structural changes while controlling graph size. We evaluate 21 detectors on the UB-GOLD benchmark and on synthetic and molecular graphs with controlled structural changes. About half of the checkpoint-selection gain is linked to choosing a checkpoint and reporting its performance on the same labeled test set. The mean gap persists across training seeds and when normal and anomalous graphs come from different datasets. Most detectors miss many tested structural changes that simple counts detect. With tuned hyperparameters, the two leaders remain the top two by AUROC under every checkpoint rule. Across nine main datasets, test-label checkpoint selection raises the mean AUROC of UB-GOLD's 12 detectors by 5.12 points relative to the final checkpoint.

cs.LG · cs.SI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37434

FeatureArc: Benchmarking Coding Agents on Evolving Features

Yunlong Wang, Alyssa Wu, Cuiyun Gao, Jia Feng, Songyuanyi Lu, Zhichao Hu

原文摘要

A software feature usually evolves over time: developers extend its first version as new use cases and requirements arrive, and each extension builds on code written earlier. Most feature benchmarks evaluate a single request on a fixed codebase. Feature evolution requires evaluating successive extensions on an agent's own earlier code and checking whether the feature remains correct. We introduce FeatureArc to evaluate coding agents on evolving features, together with a pipeline that constructs its tasks from repository histories. The benchmark combines dependencies between successive stages of the same feature, explicit requirements and interface contracts, and executable checks across four programming languages. The pipeline validates fail-to-pass (F2P) tests for new behavior and pass-to-pass (P2P) tests for existing behavior on the reference code before and after each stage. We evaluated 10 coding agents on 73 manually reviewed tasks of 2 to 10 stages. Each agent works in one persistent session and workspace without external test feedback or rollback. The evaluation reveals a gap between first-stage and whole-feature success: pooled success falls from 62.7% at the first stage to 28.5% when the whole feature is tested on the final code. Trajectory analysis identifies inherited defects as a common failure mechanism, present in 72.9% of failed later stages. These earlier errors often remain undetected: in stages following a failure, agents leave the earlier error unnoticed about 95% of the time, and their self-tests rarely revisit earlier stages. Evolving a feature autonomously, without human checks between stages, therefore remains a substantial challenge for current coding agents. At Tencent, the pipeline extends internal evaluations to this setting and guides the development of Hunyuan, Tencent's family of large language models.

cs.SE · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37432

Request-Adaptive Execution for Efficient Hybrid LLM Inference

Fahao Chen, Xuzhe Yan, Qi Zhang, Peng Li, Dongxiao Yu, Song Guo

原文摘要

Hybrid large language models (LLMs) have emerged as a promising way to combine the modeling capability of the Attention mechanism with the efficiency of state space models (SSMs). However, most existing hybrid LLMs still inherit a static design philosophy: once the Attention-to-SSM layout is chosen, every request must follow the same hybrid execution. This one-layout-fits-all design is inefficient for real-world serving workloads, where different requests rely on contextual information in different ways. This paper presents Morph, a serving system that co-designs the model architecture and runtime for request-adaptive hybrid LLM inference. Its model architecture pairs Attention and SSM layers within each block and uses prefill-oriented gating to select a request-specific layout that remains fixed throughout decoding. Its serving runtime preserves batching efficiency under divergent layer choices through selective batching and enables correct prefix reuse through hybrid-aware context caching. Experiments on six datasets show that Morph achieves a better quality-efficiency trade-off than existing static hybrid baselines. System-level evaluation further shows that Morph improves the serving throughput by up to 3.87x over baselines while substantially reducing inference latency.

cs.DC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.02466

SD-DPC: Sparse Dictionary Differentiable Predictive Control

Ali Reza Daneshvar Garmroodi, Jan Drgoňa

原文摘要

We present sparse dictionary differentiable predictive control (SD-DPC), a framework for learning sparse, interpretable feedback policies for nonlinear systems from data. A prediction model is first identified by rollout-based sparse identification of nonlinear dynamics (SINDy), building on gradient-based and multistep formulations. The policy is then parameterized as a sparse combination of dictionary functions and trained by differentiating a constrained finite-horizon predictive-control objective through this model, so that its terms are selected by closed-loop performance rather than by imitating a previously trained controller. The result is an explicit feedback law with only a handful of terms. Across three benchmark control problems, SD-DPC satisfies the constraints in almost all test scenarios, outperforms a policy distilled onto the same terms by up to an order of magnitude, and requires orders of magnitude less memory and online computation than an optimization benchmark, while admitting explicit sensitivity bounds.

eess.SY · cs.LG · math.OC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37430

When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models

Ananth K Suresh, Aveen Dayal, Gouthaman KV, Aditya Pola, Vineeth N Balasubramanian

原文摘要

For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet refusal alone is not necessarily evidence of such recognition: models may refuse because of unrelated safety constraints or without identifying the missing evidence. We study how refusal on visually unanswerable questions is shaped by three sources of control: inference-time instructions, post-training, and internal activation interventions. Across three MLLMs on the MoHoBench benchmark, we systematically manipulate the meaning and placement of refusal instructions, compare supervised and preference-based post-training, and intervene on a refusal-related activation direction derived from naturally answered and refused examples. We find that refusal is highly sensitive to both the meaning and placement of instructions, with model-specific changes at the sample level that are obscured by aggregate refusal rates. Evidence-aware and general refusal instructions have stronger effects than an intentionally mismatched safety instruction. Post-training produces larger increases in refusal, but can also induce responses that combine refusal language with attempted answers. In contrast, activation intervention produces smaller yet transferable changes across prompt conditions, without evidence of indiscriminate refusal on tested visually answerable samples. These findings show that refusal is jointly shaped by external instructions, post-training, and internal representations, highlighting the need to disentangle evidence recognition from a model's learned propensity to refuse.

cs.LG · cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37429

Bayes-Adaptive MDPs for Rollout Allocation to Enhance GRPO Training

Faeze Moradi Kalarde, Armin Toroghi, Nam Ho Koh, Doo Hwan Hwang, Seungjin Kim, Kee-Eung Kim, Scott Sanner

原文摘要

Group Relative Policy Optimization (GRPO) has demonstrated strong performance in LLM post-training. However, it allocates a uniform rollout budget to every training question, ignoring variations in question difficulty. As a result, highly difficult questions may receive too few rollouts to generate informative positive reward signals, limiting their contribution to training, while moderately difficult questions often receive more rollouts than necessary. In this work, we formulate the problem of difficulty-aware rollout budget allocation to maximize training utility under the same total rollout budget as standard GRPO. We propose AdaRoll-GRPO, which reformulates rollout allocation as a sequential decision-making problem and leverages a Bayes-adaptive Markov decision process (MDP) to learn an adaptive allocation policy. The learned policy is computed offline and applied during GRPO training without introducing additional online optimization overhead. Extensive experiments on multiple reasoning datasets and LLM architectures demonstrate that AdaRoll-GRPO consistently improves GRPO training performance under a fixed rollout budget, highlighting the effectiveness of adaptive rollout allocation for GRPO in LLM post-training.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v9arXiv:2507.04346

Action-Increment-Constrained Policy Optimization for Smooth Online Adaptive Critic Flight Control with Actuator Rate Constraints and Low Computational Complexity (extended version with Ultimate Uniform Boundedness analysis)

Yifei Li, Erik-Jan van Kampen

原文摘要

Approximate Dynamic Programming (ADP) enables online control-law adaptation through adaptive critic methods. A safety challenge in applying ADP to angle-of-attack (AoA) tracking control is that an overly sensitive neural network (NN) controller may generate oscillatory or saturated control actions under nonlinear dynamics, potentially compromising flight stability and increasing actuator stress. To address this challenge, this paper proposes two innovations: (1) the actuator rate constraints are enforced within a an Action-Increment-Constrained Policy Optimization (AI-CPO) framework inspired by safe reinforcement learning (RL), in which a primal-dual approach adaptively adjusts the penalty weight to address actuator rate constraints. This optimization is solved online using adaptive Temporally Smoothed Incremental Model-Based Heuristic Dynamic Programming (TS-IHDP) for smooth and computationally efficient online adaptive control; and (2) a low-pass filter integrated into the cascaded NN architecture to attenuate high-frequency pitch-rate commands while preserving the Markov property during online learning by exploiting the hierarchical actor structure. Compared with the IHDP baseline, simulations using a longitudinal air vehicle dynamics model demonstrate that TS-IHDP improves control smoothness and tracking performance, reduces system oscillations, and enhances the trained policy's robustness to model uncertainties.

eess.SY · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37427

Time-Dynamic Semantic-Activity Entanglement: Quantifying Communication-Based Synchrony in Collaborative VR

Pooja Pol, Jianing Zhang, Qing Huang

原文摘要

Collaborative virtual reality (VR) produces rich multimodal traces of team behavior, yet practical methods for characterizing time-varying team communication remain limited. We present time-dynamic semantic-activity entanglement (TSAE), inspired by the entanglement concept introduced by Gloor et al. (2022), to describe communication-based team synchronization within discrete time windows. TSAE is modeled by combining relative speaking activity with contextual semantic alignment for participant pairs, which is aggregated into participant-level trajectories and summarized through the Gini coefficient to understand the distribution of team-level communication synchrony. This approach is employed with German-language transcripts from fifteen collaborative VR gameplay sessions. Transcript utterances are represented using a long-context embedding model with a late-chunking strategy and analyzed within fixed game-time windows. A representative session illustrates shifting dyadic alignment and low-to-moderate participant contributions, with sporadic periods of high entanglement. Across fifteen game sessions, the metrics indicate complementary insights, offering a qualitative contrast to the core-periphery pattern reported in the foundational work. Hence, these descriptive findings support the study of communication as a dynamic process and provide a mathematically grounded approach to analyze its evolving distribution in shared interactive VR spaces.

cs.HC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.35235

False-Alarm Improvements and Detection Trade-Offs in Quantum Machine Learning for Intrusion Detection: A Reproducible Secondary Analysis

Roberto Fernández-Barrios, Iker Pastor-López, Asier González-Santocildes, Pablo García Bringas

原文摘要

Published results in quantum machine learning for intrusion detection (QML-IDS) require joint interpretation of false-positive rate (FPR), attack true-positive rate (TPR), and evaluation provenance. We reconstruct operating points in the 25-entry frame selected by Kaissar and colleagues: 21 included articles, 20 identified documentary families, substantive methodological material consulted for 15, and 24 descriptive points from five. In one test-labelled comparison, a quantum support vector machine reduces FPR from 6.875% for random forest to 3.329%, while TPR decreases from 93.955% to 93.682%; both rates improve against three other comparators. In another study, a complete benign row identifies FPR as 12/6000 (0.200%), despite incomplete attack-side allocation that requires additional assumptions for TPR. Normalized matrices identify detection ranges rather than a unique aggregate rate. A uniform provenance rule separates training-described and test-adjusted results from six conditional illustrations whose independence remains unverified. The contribution is a reproducible interpretation of favorable comparisons and detection trade-offs, without converting descriptive results or hypothetical calculations into operational guarantees.

cs.CR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37424

Privacy-Preserving Over-the-Air Federated Learning over Cell-Free Massive MIMO System with User-Centric Clustering

Manobendu Sarker, Soumaya Cherkaoui

原文摘要

This paper investigates cross-layer design for differentially private federated learning over cell-free massive MIMO networks with over-the-air (OTA) aggregation. Distributed access points (APs) collaboratively serve user equipments (UEs) through user-centric clustering, where clipped local model updates are perturbed with Gaussian noise and simultaneously transmitted over a shared uplink for OTA averaging at a central processing unit. Exploiting channel hardening, we show that thermal noise inherent in wireless reception can act as a privacy-masking resource, reducing or even eliminating the need for artificial noise. Based on this observation, we formulate a joint optimization problem over AP-UE associations, UE transmit powers, and artificial noise variances to minimize aggregation variance subject to per-UE zero-concentrated differential privacy constraints derived from an ergodic privacy accounting framework. Analysis reveals a regime-dependent structure: stringent privacy requirements favor low transmit powers and sparse AP associations to exploit thermal noise for privacy, whereas relaxed requirements favor larger clusters to improve aggregation quality. The resulting mixed-integer nonconvex problem is solved via alternating optimization combining difference-of-convex power allocation and greedy coordinate-descent clustering. Numerical results show up to 89% lower aggregation variance than conventional baselines and enable 97% test accuracy on MNIST with consistent gains on CIFAR-10. Under strict privacy budgets, competing schemes remain near random-chance performance, while under relaxed budgets they continue to lag significantly behind the proposed design.

cs.IT · eess.SY

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37422

Beyond the Readout: A Latent Numerical Frame Shared Across Modality and Operation

Oriol Juan Sabater, Alex Ferrando de las Morenas, Pau Rodríguez, Xavier Suau

原文摘要

When a multimodal model computes 17+35 from text or 91-39 from a rendered image, it produces the same answer, 52. Does it use one internal concept of number for both, or separate machinery for each path? Similarity metrics such as RSA or CKA cannot settle this question. They give the same score to identical and to rotated directions, and two representations can look alike only because both prepare the same output digits. We therefore use causal interventions. For four conditions (text or image, addition or subtraction), we learn a small subspace of Gemma4-12B-it's residual stream in which swapping coordinates swaps the model's answer. The subspaces answer three questions. First, do the conditions write numbers along the same directions? They do not. Reusing a subspace from text in an image on the same operation recovers only 53.2% of its causal effect. A single rotation and rescaling maps one subspace onto the other and raises recovery to 91.4%, so the conditions share one geometry expressed in different coordinates. Second, does this shared geometry only reflect the digit the model is about to print? To test this, we split each subspace into 9 directions that the output layer reads to choose the next digit and 13 directions that it cannot see. The 13 hidden directions still carry the number, yet changing them barely alters the digit being written. Third, what are these hidden directions for? A model must emit a multi-digit answer one digit at a time, so it can compute the full result early and keep it in reserve. When two numbers share their first digit, swapping the number in the hidden directions leaves the digit being written intact and redirects the next digit in 77.6% of cases. The model thus forms an abstract numerical concept upstream and unfolds it across time: multimodal representations share meaning through relational geometry across distinct coordinate frames.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2604.14809

Expert-Guided Class-Conditional Goodness-of-Fit Scores for Interpretable Classification with Informative Missingness

Shahar Cohen, David M. Steinberg, Yael Radzyner, Yochai Ben Horin

原文摘要

We study a classification problem with three key challenges: pervasive informative missingness, the integration of partial prior expert knowledge into the learning process, and the need for interpretable decision rules. We propose a framework that encodes prior knowledge through an expert-guided class-conditional model for one or more classes, and use this model to construct a small set of interpretable goodness-of-fit features. The features quantify how well the observed data agree with the expert model, isolating the contributions of different aspects of the data, including both observed and missing components. These features are combined with a few transparent auxiliary summaries in a simple discriminative classifier, resulting in a decision rule that is easy to inspect and justify. We develop and apply the framework in the context of seismic monitoring used to assess compliance with the Comprehensive Nuclear-Test-Ban Treaty. We show that the method has strong potential as a transparent screening tool, reducing workload for expert analysts. A simulation designed to isolate the contribution of the proposed framework shows that this interpretable expert-guided method can even outperform strong standard machine-learning classifiers, particularly when training samples are small.

stat.ML · cs.LG · stat.AP

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.24764

BanglaBBQ: A Bias Benchmark for Question Answering in the Bangladeshi Context

Sumit Anthony Gomes, Yasin Rahman, Taeeba Tasnia, Faiza Nooren Suhita

原文摘要

This paper presents BanglaBBQ, a question-answering bias benchmark designed to capture stereotypes grounded in the Bangladeshi context, in both Bangla and English. Following KoBBQ and PakBBQ, BanglaBBQ adapts the original Bias Benchmark for Question Answering (BBQ) by classifying its templates into three groups: directly translated (DT), modified to align target groups or contexts with Bangladesh (TM), and removed when deemed unsuitable for the Bangladeshi context (SR). This process retains 140 templates across six categories, comprising 76 DT and 64 TM templates. Five newly authored (NA) categories, based on stereotypes common in Bangladeshi society, are then added: marital status, age and political perception, regional origin, educational background, and mental-health status. Native Bangladeshi annotators judged all 466 newly authored templates, and 372 were retained because a strict majority of English annotators identified the intended target and rated the attitude as common enough. Each newly authored template is linked to a documented source and a verbatim quote from it. Each evaluated template also has a counter-stereotypical version of its disambiguated context. The dataset was evaluated using two open-weight models (Qwen3-4B and Qwen3-8B) and two closed models (GPT-4.1-mini and Gemini 2.5 Flash). All four lean towards the stereotyped person when the context is ambiguous, in both languages, and the effect of language differs by model, so English results do not predict behaviour in Bangla. On the adapted BBQ templates, both Qwen3 models are also less accurate when the context contradicts a stereotype than when it confirms it. On the newly authored templates, the closed models follow counter-stereotypical evidence almost perfectly, and all four models are more biased on the templates that annotators validated than on those they rejected.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2503.04956

Foreclassing: From Forecasts to Decisions

Daniel Andrew Coulson, Martin T. Wells

原文摘要

Time series forecasts are widely used to inform decisions under uncertainty. In many applications, these decisions involve classifying a time series based on outcomes that have yet to be observed. In this paper, we introduce Foreclassing, a new machine learning problem in which the class label of a time series is defined as a function of its future observations. We provide a mathematical formulation of Foreclassing and establish several theoretical results characterizing the problem. These results describe how additional future information affects Bayes classification risk, provide bounds on the classification error of learned Foreclassing models, and demonstrate how future observations used during training can yield finite-sample improvements over conventional classifiers. To provide a benchmark for future research, we propose ForeClassNet, a family of neural network models that jointly perform forecasting and classification while accounting for uncertainty in future outcomes. We investigate a variety of model architectures and evaluate their performance against conventional time series classifiers and forecasting-then-classify approaches. Experiments on datasets spanning weather, energy, and finance demonstrate the potential of Foreclassing to improve classification performance, probabilistic prediction, and downstream decision-making. Together, these theoretical and empirical contributions establish a foundation for future research on Foreclassing.

stat.ML · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37418

Neighborhood Deprivation Shapes Post-Flood Population Losses in Pune and Bandung

Sukie Xiuqi Yang, Guangyuan Weng, Shiyue Cui, Zhicong Chen, Catherine Chen

原文摘要

Disaster research in high-income settings shows that better-resourced communities see population inflows and faster recovery, while disadvantaged communities suffer prolonged population loss. Whether this pattern holds in rapidly urbanizing cities of the Global South remains unclear. Here we show, using mobile phone data from Pune, India, and Bandung, Indonesia, how neighborhood-level changes in phone presence, a proxy for short-term population movement, relate to flooding and neighborhood deprivation. We combine satellite-derived flood maps with phone records to compare flooded and non-flooded areas over the 30 days around each flood. In both cities, less deprived neighborhoods see phone presence rise after flooding, while more deprived neighborhoods see it fall, consistent with resource-seeking interpretations documented in North American and European disaster research. The cities differed, however, in how deprivation relates to flood exposure: in Pune, more deprived flooded neighborhoods showed suggestive evidence of disproportionately larger decreases in detected device presence than similarly deprived non-flooded neighborhoods; no comparable interaction was observed in Bandung. Economic considerations are associated with post-flood device-presence changes in South and Southeast Asian cities in ways that parallel high-income contexts, though local context adds variation.

cs.CY

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.30058

HEART-Bench: Do LLMs Make Persona-Consistent Decisions?

Weihan Peng, Chenxu Zhang, Qianao Wang, Yuling Shi, Heng Lian, Qihong Mao, Jiahao Pang, Chunliang Feng 等 10 位作者

原文摘要

LLM agents in interactive entertainment and social applications, including companions, role-play characters, and persona clones, need decisions grounded in personality and past experiences. We introduce HEART-Bench to evaluate memory-grounded persona-consistent response selection. The benchmark combines 11 Big Five-anchored synthetic characters, each with 1,000 autobiographical memories, and 64 DIAMONDS-guided decision scenarios into 673 multiple-choice questions. Domain experts independently compare candidate responses, with a senior expert resolving disagreements and revising responses when needed; other characters' adjudicated responses to the same scenario serve as distractors. In the main evaluation, models receive character IDs and occupations, scenarios, and options, with retrieved memories added in memory-augmented settings. Detailed personality-and-value profiles are withheld. We evaluate 16 LLMs under No-Retrieval, Naive-RAG, Mem0, and PersonaDB. Accuracy without memories stays near the 25% chance level, and the best memory-augmented configuration reaches 64.9%. On a matched 150-item subset, human annotators reach 80.8%, compared with 64.0% for the best model. Memory-content ablations show the importance of retrieving experiences that match both the target character and the decision context.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37415

The Envelope Layer: An Underestimated Surface of LLM Agent Defense

Yuelin Han

原文摘要

Prior work framed the envelope layer as the channel through which malicious content enters an agent, and established it as a defense surface independent of the LLM layer and the agent layer; however, that work only tested the envelope layer briefly and did not examine how label strength, envelope authenticity, scaffold differences, or existing memories affect defense effectiveness. Following its GDA Measurement methodology and the A/B/C/D outcome taxonomy, this paper takes the A2A-TIBA attack as the test vehicle, uses the semantic breach rate as the primary indicator with the semantic refusal rate as its complementary reference, and characterizes the security boundary of the envelope layer through a two-by-two factorial experiment and a memory experiment. The results show that: when the envelope is authentic, strong labels consistently reduce the semantic breach rate, whereas weak labels are not necessarily better than an unlabeled baseline; when the envelope can be forged, the benefit of strong labels is weakened and its direction varies across frontends; and memory has a larger effect than any label manipulation: success memory raises the semantic breach rate while failure memory lowers it, and most combinations that contain both experiences return to the no-memory baseline. Accordingly, this paper refines the envelope layer from a single admission channel into two levels, an admission channel plus a context assembly surface, and argues that its security boundary is not a fixed property of labels but a dynamic outcome jointly shaped by label claims, frontend permissions, and historical memory.

cs.CR

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2604.18215

Memorize When Needed: Decoupled Memory Control for Spatially Consistent Long-Horizon Video Generation

Yanjun Guo, Zhengqiang Zhang, Pengfei Wang, Xinyue Liang, Zhiyuan Ma, Lei Zhang

原文摘要

Spatially consistent long-horizon video generation aims to maintain temporal and spatial consistency along predefined camera trajectories. Existing methods mostly entangle memory modeling with video generation, leading to inconsistent content during scene revisits and diminished generative capacity when exploring novel regions, even when trained on extensive annotated data. To address these limitations, we propose a decoupled framework that separates memory conditioning from generation. Our approach significantly reduces training costs while simultaneously enhancing spatial consistency and preserving the generative capacity for novel scene exploration. Specifically, we employ a lightweight, independent memory branch to learn precise spatial consistency from historical observations. We first introduce a hybrid memory representation to capture complementary temporal and spatial cues from generated frames, then leverage a per-frame cross-attention mechanism to ensure each frame is conditioned exclusively on the most spatially relevant historical information, which is injected into the generative model to ensure spatial consistency. When generating new scenes, a camera-aware gating mechanism is proposed to mediate the interaction between memory and generation modules, enabling memory conditioning only when meaningful historical references exist. Compared with the existing method, our method is highly data-efficient, yet the experiments demonstrate that our approach achieves state-of-the-art performance in terms of both visual quality and spatial consistency. Project page can be found at https://iguoyanjun.github.io/memorize-when-needed-project-page/.

cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37413

Displacement Moments and Path-Terminal Separation in Brownian Interpolant Sampling

Zhong Han

原文摘要

Brownian stochastic interpolants connect paired endpoint samples, yet their sampling error can depend on displacement moments even when neither endpoint has a first moment. For arbitrary Borel couplings, we prove early-stopped, target-to-output KL guarantees for standard Euler sampling from the true source. A fourth displacement moment gives an explicit $O(N^{-1})$ bound; a second moment suffices for fixed-law convergence as drift error vanishes. For one fixed coupling with no displacement moment above two, original bridge-path KL stays of order one, comparison-path KL is of order $1/\log N$, and terminal KL is at most $C/N^2$ on the same grid. The path decomposition identifies the information about the source discarded by Markov projection. When the original bridge is Markov, current-state Euler sampling instead has a second-moment $O((M_2+d)/N)$ bound; finite-entropy Schroedinger bridges for the probability Brownian reference are included. We also construct a current-state learner with expected output KL consistency from iid endpoint pairs. Numerical diagnostics illustrate the distinct errors. Conditional diffusion identities specify a restoration interface, while image controls find no iterative advantage over cheap probabilistic predictors. Uniform terminal rates for general second-moment couplings remain open.

cs.LG · math.PR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37412

The Capacity of Private Information Retrieval from Coded Storage Systems with Colluding, Robust and Byzantine Servers

Rui Sun, Yiwei Zhang

原文摘要

We consider the problem of private information retrieval from MDS-coded databases with $T$ colluding servers, $U$ robust servers, and $B$ Byzantine servers, referred to as U-B-MDS-TPIR. In the U-B-MDS-TPIR setting, $M$ files are stored across $N$ servers, and in this paper each file is independently encoded using an $(N,K)$ GRS code. We characterize the capacity among linear, full support-rank U-B-MDS-TPIR schemes as $C=\frac{N-U-2B}{N-U}(1+\frac{K+T-1}{N-U-2B}+\cdots+(\frac{K+T-1}{N-U-2B})^{M-1})^{-1}$ when $N-U-2B\ge KT$. When $K=1$ or $T=1$, the same expression gives the capacity without the linearity and full support-rank restrictions. We further extend the construction to the retrieval of $P$ out of $M$ files. For $P \ge \frac{M}{2}$, the resulting multi-file scheme achieves capacity when $K=1$ or $T=1$.

cs.IT

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37738

Poisoned Skills: Propagating Jailbreaks Through Agent Markdown Files

Kashyap R Manjusha, Murali Emani

原文摘要

Modern agent frameworks allow users to provide Markdown files containing system prompts, tool descriptions, and skill definitions (e.g., SKILL.md), which are incorporated into the agent's context during initialization. In this work, we investigate whether this configuration mechanism can also serve as an injection channel: can a jailbreak embedded in a Markdown skill file propagate through an agent and induce prohibited behavior? To answer this question, we construct a LangChain-based question-answering agent whose system prompt is assembled from a user-provided Markdown file. We embed a jailbreak combining persona manipulation with a GPTFUZZER-based attack and evaluate its effectiveness across seven contemporary open-weight models using 270 forbidden questions spanning nine harm categories. Under a benign system prompt, the models exhibit near-complete refusal behavior, with a mean attack success rate (ASR) of just 0.5%. In contrast, the injected skill file raises the mean ASR to 45%, reaching 68.9% for individual models. We further evaluate two payload obfuscation strategies: an invisible zero-width character-based stealth payload and a phonetic payload based on grapheme-to-phoneme transformations. Neither outperforms the plain-text injection; notably, the phonetic variant reduces mean ASR to 30%. These findings identify Markdown skill files as an underexplored attack surface for AI agents and suggest that vulnerability to configuration-based jailbreaks depends more strongly on the underlying model than on the specific payload obfuscation technique.

cs.CR · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37409

Can Complementary Signals Bridge Similarity Islands? Manifold-Augmented Graph Embedding for Multimodal Recommendation

Zhenye Yang, Jinpeng Chen, Wenbo Fu, Jun Ma, Huan Li, Qimin Zhou, Tao Wang, Li Sun 等 9 位作者

原文摘要

Multimodal Recommender Systems (MMRS) leverage rich item features to mitigate data sparsity, yet similarity-driven graph construction and propagation can favor semantically similar substitutes over functionally related complements. When behavioral associations connect semantically distant items, complementary signals can remain geometrically isolated, yielding fragmented "Similarity Islands" and semantic drift when standard propagation is applied. To overcome these limitations, we propose MAGE (Manifold-Augmented Graph Embedding), a framework for personalized multimodal ranking that decouples complementarity modeling into offline geometric warping and online topological propagation. The offline phase enhances item representations by extracting complement-aware residual directions via a Soft Conceptor mechanism and injecting them under a local gap-preserving constraint, incorporating co-purchase evidence while controlling the magnitude of geometric perturbation. Synergistically, the online phase employs an asymmetric dual-stream GNN that treats visual features as stable anchors and augmented textual features as flexible probes, fused through a self-calibrated cross-modal mechanism to reconcile shared semantics and modality-specific residual cues. Extensive experiments on four Amazon datasets show that MAGE yields consistent improvements over strong multimodal recommendation baselines. Together with the supporting analyses in the appendix, these results suggest that geometry-aware augmentation can alleviate the semantic mismatch that arises in complementary recommendation settings. Our code is available at https://github.com/Alex-bupt/mage.

cs.IR

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.09552

Phases of Muon: When Muon Eclipses SignSGD

Elliot Paquette, Noah Marshall, Lucas Benigni, Guangyuan Wang, Atish Agarwala, Courtney Paquette

原文摘要

Muon and related spectral optimizers have recently demonstrated strong empirical performance as scalable stochastic methods, often outperforming Adam. Yet their behavior remains poorly understood. We analyze stochastic Muon, which orthogonalizes an exponential moving average of minibatch gradients, on a high-dimensional matrix-valued least squares problem, and we compare it with stochastic SignSGD, a simple baseline comparable to Adam. We derive explicit deterministic dynamics that provide a tractable framework for studying learning behavior. We show that Muon has non-linear sensitivity to the effective batch $B_{\rm eff}=B(1+β)/(1-β)$. With exact whitening and small $B_{\rm eff}$ Muon fails to resolve smaller eigenmodes, slowing down convergence. Momentum and finite Newton--Shulz iterations can act as a ridge which lessen this effect. This behavior is in contrast with SignSGD which has no transition, leading to different optimal learning rates, but with matching convergence rate up to a constant factor in the isotropic setting. For anisotropic data, large batch setting Muon performs a square-root preconditioning with respect to the data covariance spectrum while SignSGD provides no preconditioning for generic covariances. We analyze time to target loss in a power law covariance model and derive a phase diagram in the data exponent $α$ and target exponent $ζ$. Muon is favored for steep covariance spectra with rich targets ($α>2$, $ζ<α+1$), SignSGD is favored for flat spectra ($α<1$), and in other regions the winner depends on how the target loss scales with dimension. Our results shed light on why replacing Adam with Muon is not trivial even from a pure optimization perspective.

math.OC · cs.LG · stat.ML

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37406

CTBakeGS: training-free baking of photorealistic CT volume renderings into 3D Gaussian splats

Haruki Fukuda, Constantin Lux, Wolf Schweitzer, Tadashi Hosoya, Lars Ebert

原文摘要

Photorealistic volume rendering of computed tomography (CT) may convey anatomy intuitively, but it needs the CT data, dedicated software and a powerful graphics processor. Three-dimensional Gaussian splatting (3DGS) displays photorealistic scenes in real time on less powerful devices, but existing medical approaches train each scene for minutes to hours. We aimed to convert a CT volume rendering directly into a 3DGS model without training. Splats were placed on an adaptive octree of millimetre-sized cells. Their colour and opacity were computed on the graphics processor with the renderer's lighting, cell opacity was converted analytically into splat opacity, and faint splats were retained stochastically above the 1/255 opacity floor of standard 3DGS rasterisers. Six public CT scans (1.5-mm voxels) were converted with three transfer functions and compared with converged references by peak signal-to-noise ratio (PSNR), structural similarity and a learned perceptual metric. Two-million-splat models reached a PSNR of 32.6 +/- 0.9 dB for bone, 30.1 +/- 1.0 dB for contrast-enhanced vessels and 32.3 +/- 1.5 dB for skin and bone. A conversion pass (usually two or three per model) took 4-30 s on a current laptop and 9-60 s on a desktop with a 4-GB graphics card. On the laptop, the models rendered at 68 frames per second at 800 x 800 pixels. Training-free conversion produced real-time 3DGS models of photorealistic CT renderings in about a minute or less per conversion pass on ordinary computers, with fidelity that we consider sufficient for interactive viewing.

cs.CV · cs.GR

查看论文
arXiv ↗PDF ↗