跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1680新增论文
+1333其中 AI 相关
+187已有论文修订
+59新增 AI 导读

arXiv 最近完成扫描:2026-11-18 04:28 · 全站最近入库:2026-11-17 23:35各信源进度 →

统计口径与数据新鲜度

统计 2026-11-17 08:57 — 2026-11-18 08:57(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:2026-11-18 04:28最近发现新增版本:2026-11-17 23:35库内论文最新更新时间:2026-11-17 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。

25,786 已收录论文21,491 AI 相关7,725 有修订112 HF 社区发现

浏览论文库

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 5,970 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-18 04:27(北京时间)
新论文 · v1arXiv:2611.19975

Federated Dataset Simulation: Inducing Label-Free Heterogeneity Across Tasks

Vasilis Siomos, Lam Ngo, Jonathan Passerat-Palmbach, Giacomo Tarroni

原文摘要

Federated Learning enables collaborative model training across decentralized clients without sharing raw data. While the field has seen rapid growth, the datasets underpinning FL evaluation have not kept pace: real federated datasets remain scarce, and most studies rely on small-scale centralized classification datasets partitioned via Dirichlet label skew, a one-dimensional proxy for the multi-dimensional heterogeneity present in real federations. Moreover, label-based partitioning is by construction inapplicable to non-classification tasks such as segmentation and detection, for which no principled simulation method exists. We cast the conversion of centralized datasets into federated benchmarks as the task of Federated Dataset Simulation (FDS), and decompose it into a representation function $Φ$ and an assignment mechanism $π$. We provide two label-free instantiations -- geometric partitioning via pretrained model embeddings and semantic partitioning via vision-language model-derived criteria -- applicable to classification and dense prediction tasks alike. We evaluate across four datasets spanning classification, segmentation, and detection, and show that both methods induce heterogeneity that matches or exceeds that of label-skew partitioning, with measurable impact on downstream FL performance. We release fds, a modular open-source package for end-to-end federated dataset simulation, heterogeneity measurement, and FL training, at https://github.com/siomvas/fds.

cs.LG · cs.AI

版本更新 · v2arXiv:2609.05327

LLM-Driven Algorithm Design for Quantum Circuit Synthesis based on Binary Decision Diagrams

Yoonju Sim, Federico Berto, Chuanbo Hua, Jinkyoo Park, Changhyun Kwon

原文摘要

Quantum circuits are central to implementing quantum algorithms on quantum devices, where quantum gates must be reversible. Many quantum algorithms rely on Boolean functions, which must therefore be implemented reversibly within quantum circuits. Reversible circuit synthesis provides a way to translate such Boolean functions into reversible circuits. Binary decision diagrams (BDDs) offer a scalable approach to this task, but the resulting BDDs and circuits depend heavily on variable ordering. Existing ordering heuristics commonly minimize BDD size because it is closely tied to the circuit size. However, BDD size is an imperfect proxy for the quantum cost of the synthesized circuit (QCC). We propose \texttt{QuantumEvo}, an evolutionary framework that uses an LLM as a heuristic generator for QCC-aware BDD variable ordering. Instead of predicting orderings directly, \texttt{QuantumEvo} searches over ordering heuristics initialized from multiple heuristic families. Candidate heuristics directly manipulate variable orderings using standard BDD operations and are selected by downstream QCC. The discovered heuristic, HGA-QE, modifies the sifting step inside a genetic algorithm so that the procedure is better aligned with QCC. Across the benchmark set, HGA-QE achieves a 70.9\% tie-or-win rate against the per-function best baseline and is strictly best on 13.5\% of the functions. The results demonstrate broadly competitive QCC performance, with HGA-QE showing a clearer relative advantage in strict wins on the two benchmark suites drawn from sources different from the data used for heuristic discovery.

cs.AI · cs.AR

版本更新 · v2arXiv:2507.14121

Are Kolmogorov-Arnold Networks (KANs) Suitable for Imbalanced Data Classification?

Pankaj Yadav, Vivek Vijay

原文摘要

Kolmogorov-Arnold Networks (KANs) parameterize each connection as a learnable univariate B-spline, offering native feature-level interpretability. Whether this locality helps or hinders under class imbalance, where minority sparsity may leave spline edges without enough local observations to update reliably, remains unexamined. We evaluate KANs against capacity-matched Multilayer Perceptrons (MLPs) and XGBoost across ten imbalanced tabular datasets under three conditions: native class distribution, Focal Loss, and SMOTE-Tomek resampling, within a parameter-parity, two-phase framework. Under the native distribution, KAN spline activations collapse in minority-sparse regions, showing stronger majority-class bias than the matched MLP. Focal Loss yields negligible improvement, as it reweights gradients without providing the localized observations needed for discriminative curvature. Conversely, SMOTE-Tomek resampling lets spline activations recover structured minority-class representations, achieving the best cohort-wide rank and a large, statistically significant gain over unmitigated KANs, though not over resampled baselines. KAN's native spline interpretability directly reveals these dynamics via learned activation profiles, proving representational change rather than threshold recalibration without post-hoc attribution. However, these interpretability gains incur substantially higher training overhead than a parameter-matched MLP across all conditions, with no strategy-dependent latency reduction. Thus, KANs are conditionally suitable for imbalanced tabular classification: effective and interpretable with data-level resampling, yet costly relative to simpler alternatives without it.

cs.LG · cs.AI

新论文 · v1arXiv:2611.19969

Value and Policy Calibration from Execution Feedback for Socially Compatible Driving

Sen Shen, Ning Liu, Xiangrui Kong, Yingwei Li, Dongkun Han

原文摘要

Autonomous vehicle (AV) decisions at intersections affect the delay, braking, and motion smoothness of surrounding human-driven vehicles (HDVs). Learning socially compatible behavior requires connecting imagined interaction outcomes with feedback from executed actions. We propose a world-model reinforcement learning framework combines imagined policy improvement with direct value and policy calibration from finalized interactions, balancing AV progress with HDV interaction burden. An interaction-aware latent model supports imagination, while off-policy correction and replay updates enable observed rewards and transitions to supervise both the critic and actor. Simulation experiments at an unsignalized intersection achieve a 100% success and reductions in following-HDV acceleration variation of 65.5% and 44.0% relative to PIWM and DreamerV3, respectively. Ablations identify complementary roles for value calibration in reducing HDV burden and policy calibration in improving passage efficiency. Tests under out-of-distribution traffic shifts further show lower ego jerk and HDV acceleration variation than PIWM at comparable passage times.

cs.RO · cs.AI

版本更新 · v2arXiv:2605.17620

SynVA: A Modular Toolkit for Vessel Generation and Aneurysm Editing

Marten J. Finck, Niklas C. Koser, Sarker M. Mahfuz, Tameem Jahangir, Jon E. Wilhelm, Daniel Behme, Naomi Larsen, Wojtek Palubicki 等 10 位作者

原文摘要

Intracranial aneurysms (IAs), characterized by unpredictable growth and risk of rupture, are a major cause of stroke and can lead to life-threatening hemorrhages with high mortality and long-term disability. With aging populations, the incidence and overall burden of cerebrovascular diseases are expected to increase, highlighting the need for scalable approaches to analyze complex medical data and improve population-level understanding of these conditions. While digital twins and deep learning offer promising avenues for improving diagnosis, prognosis, and treatment, their effectiveness is limited by the scarcity of large-scale, high-quality medical data and corresponding labels. We present Synthetic VAsculature (SynVA), a modular toolkit for vascular mesh generation and anatomically consistent aneurysm synthesis. SynVA combines novel flow-matching-based methods for generating healthy vessel meshes with learning-based approaches for anatomy-conditioned aneurysm mesh generation; it computes aneurysms from pre-existing vascular geometries rather than generating them in isolation. In addition, we introduce the SynVA procedural model for vascular and aneurysm synthesis based solely on physiological principles and statistical priors, which enables the generation of large-scale datasets. To this end, we release a dataset of 50,000 fully labeled mesh samples for a variety of downstream vision tasks. Extensive quantitative and qualitative evaluations show that SynVA generates realistic vessel geometries and anatomically plausible aneurysms, with only minor trade-offs between expert-perceived realism and similarity to real aneurysms. Moreover, pretraining on SynVA data improves point-cloud segmentation and morphological parameter estimation on real data. Code will be released upon acceptance.

cs.CV · cs.AI · cs.LG

新论文 · v1arXiv:2611.19955

Equal Dose, Different Coverage:Tracing Rate and Duration in Image Editing

Yi Liu

原文摘要

Benchmarks for scientific image editing test whether a model draws the outcome of a described process, leaving open which representations carry its errors and whether correcting them yields the right quantity. We study this question with an airbrush of fixed deposition profile, under which exchanging rate and duration preserves the dose and should preserve the endpoint, yet changes the edited coverage. Matched-norm activation patches across text positions and denoising stages show that, in distilled FLUX with Chinese prompts, representations at unchanged text positions (context) carry this discrepancy: early context transfers 3.51 of the 3.98 percentage-point (pp) signed route difference, against 0.09 pp for equal-norm updates at the two changed digits. Early image latents transmit native (unpatched) route differences across two FLUX configurations and OmniGen, while text-position effects depend on configuration. The trace predicts an intervention: share early context between routes while each retains its own digits. On independent English layouts, this reduces the distilled-FLUX absolute route gap from 1.13 to 0.60 pp, outperforms early digit sharing, and improves agreement more than in base FLUX under directly paired tests. Calibrated evaluation, which prescribes an area for each dose, separates this gain from accuracy: sharing the full prompt encoding produces identical route images with target errors of 8-16 pp, and stating the target percentage directly retains a compressed area response. Localization predicts a consistency gain on new layouts, and calibration measures the quantity error that agreement leaves unresolved.

eess.IV · cs.AI · cs.MM

版本更新 · v5arXiv:2604.19139

Verbal tics in frontier language models: A critical review of current releases, research evidence, and public discussion

Shuai Wu, Xue Li, Zhijun Wang, Bolun Liu, Weilin Cai, Zihao Su, Ran Wang

原文摘要

Repeated praise, canned reassurance, formulaic contrasts, and conspicuous vocabulary figure prominently in discussions of large language models. Context governs how these habits should be read. A conventional phrase can orient a reader, whereas a fluent answer can reinforce a false belief. This critical review examines linguistic habits and sycophancy across eight developer families. We verify public offerings from OpenAI, Anthropic, Google DeepMind, xAI, ByteDance, Moonshot AI, DeepSeek, and Xiaomi against official release and API documentation through 1 October 2026. We then synthesize empirical research on lexical overrepresentation, stylistic variation, social warmth, and agreement alongside benchmark methods and dated public discussions in English and Chinese. Published studies document recurring linguistic patterns and agreement that distorts human judgment, but comparable measurements for the newest releases remain sparse in the retrieved material. Current user accounts report both persistent complaints and clearer prose, and experiences vary by task and prompt design. We propose separate measures for recurrence, contextual appropriateness, and belief distortion paired with exact service records and language-specific annotation. This framework keeps claims about writing quality and conversational reliability testable as model services change.

cs.CL · cs.AI

版本更新 · v3arXiv:2610.33012

Recursive Semantic Divergence for LLM Agent Consistency

Harshavardhan Abichandani, Penny Chong, Atin Ghosh, Daniel Dahlmeier

原文摘要

Large Language Model (LLM) agents with tool use can behave inconsistently across independent runs from identical states. They may call different tools, produce contradictory outputs, or follow divergent strategies. This variability compounds over time and undermines deployment reliability. Existing evaluation metrics primarily measure task correctness via completion-based scores and do not capture consistency in agent behavior across executions. Separately, agent behavior is analyzed either at the token level using probability distributions or at the single-response level using semantic clustering, but neither captures how inconsistency propagates across future turns of a trajectory. We introduce Recursive Semantic Divergence (RSD), a trajectory-level consistency metric adapted from bisimulation metrics in reinforcement learning. RSD measures the expected cumulative semantic disagreement between two independent rollouts of the same agent from the same state, using natural language inference to detect logical inconsistency. It is defined recursively over future states and approximated by a neural network trained on independent rollouts. The learned metric estimates future divergence from the current state, providing a consistency score at each turn without requiring the full trajectory to complete. When used as a reward signal for fine-tuning, RSD reduces AUC coefficient of variation (CV) by up to 30.9% on tau2-bench and 50.5% on ToolSandbox. Across Qwen-2.5 and Qwen-3.5 models, RSD reduces AUC CV in seven out of eight model-benchmark settings, while improving the final task progress in all eight settings.

cs.CL · cs.AI · cs.LG

版本更新 · v2arXiv:2611.14106

ProofCoder: Continuous Reasoning for Automated Software Verification

Martim Afonso, Inês Lynce, Maurizio Morisio, João F. Ferreira

原文摘要

Formal verification using proof assistants provides strong correctness guarantees for software, but constructing proofs remains labor-intensive and requires significant expertise. Recent approaches based on large language models (LLMs) have shown promise in automating proof synthesis. However, these approaches typically constrain the model to a rigid tactic-generation loop, in which the surrounding system controls premise retrieval, proof-state management, and backtracking. We introduce ProofCoder, an agentic proof-synthesis system for Rocq that approaches theorem proving from a coding-agent perspective. Rather than generating tactics within a fixed loop, ProofCoder reasons continuously throughout proof search and uses proof-assistant feedback to decide when to gather additional information, inspect proof states, revise previously generated proof fragments, or backtrack to an earlier point in the proof. This allows it to iteratively construct, repair, and refine proofs, at either tactic-level or whole-proof granularity. We evaluate ProofCoder on 280 challenging verification tasks from the AutoRocq Benchmark. On the best-performing configurations, ProofCoder increases proof success from 47% to 60% over the state-of-theart AutoRocq system, corresponding to a 27.7% relative improvement at approximately half the cost. Our ablations identify continuous reasoning as a key design choice: without it, the agent solves fewer lemmas at substantially higher cost, backtracks four to six times more often, and more frequently discards its own revisions. Whole-proof and tactic-level generation achieve similar success rates but follow distinct search and repair strategies. These results suggest that effective automated software verification requires more than improved tactic selection: continuous reasoning and feedback-guided proof revision are critical ingredients for robust proof synthesis.

cs.SE · cs.AI

新论文 · v1arXiv:2611.19925

Learning When to Continue and When to Stop: Termination-Aware OPSD

Chaoqun He, Wannian Yi, Changyi Sun, Chaojun Xiao, Lijie Wen

原文摘要

On-Policy Self-Distillation (OPSD) improves language model reasoning by distilling from a frozen teacher initialized identically to the student and conditioned on privileged information. However, we identify an underexplored termination mismatch in OPSD: the student fails to reliably learn when to continue and when to stop. This mismatch manifests in two opposing failure modes-premature termination, where reasoning stops before sufficient progress is made, and repetition-driven over-generation, where the model continues generating after reasoning has stalled. Our analysis shows that the teacher-student discrepancy on EOS grows during OPSD training, while length-capped trajectories are frequently dominated by unproductive repetition. To address this issue, we propose TA-OPSD (Termination-Aware On-Policy Self-Distillation), which augments the original OPSD objective with two targeted objectives. Stop/Continue Distillation addresses premature termination by explicitly aligning the teacher and student on whether reasoning should continue, while Repetition-Aware Unlikelihood Training addresses repetition-driven over-generation by discouraging repetitive continuations in the student's failed trajectories. The resulting objective improves termination behavior from complementary directions: continuing when further reasoning is needed and stopping when reasoning becomes unproductive. Experiments on mathematical reasoning tasks with Qwen3 (1.7B/4B/8B) show that TA-OPSD consistently outperforms OPSD and other baselines across all three model scales. Beyond the performance gains, ablation and behavioral analyses show that the two proposed objectives effectively mitigate their corresponding termination failures, reducing both premature termination and repetition-driven over-generation.

cs.LG · cs.AI · cs.CL

版本更新 · v3arXiv:2406.05364

Could Mobile AI be Exploited?: Understanding the Exploitations of Trust and Ethics Gap in On-Device Small Language Models

Kalyan Nakka, Jimmy Dani, Nitesh Saxena

原文摘要

In this paper, we present a very first study to investigate trust and ethical implications of on-device artificial intelligence (AI), focusing on small language models (SLMs) amenable for personal devices like smartphones. While on-device SLMs promise enhanced privacy, reduced latency, and improved user experience compared to cloud-based services, we posit that they might also introduce significant risks and vulnerabilities compared to their on-server counterparts. As part of our study, we conduct a systematic evaluation of the state-of-the-art on-devices SLMs, compared to their on-server counterparts, using well-established trust and ethics frameworks. Our results illustrate that on-device SLMs are significantly less trustworthy and lacks ethical safeguards, specifically demonstrating more stereotypical, unfair, privacy-breaching and harmful content generation behavior. Further, these risks of on-device SLMs could be exploited through Open Knowledge (OkNow) Attacks, and we demonstrated them using potentially unethical vanilla prompts, to which the on-device SLMs answer with valid responses without any filters and without the need for any jailbreaking or prompt engineering. These responses can be abused for various harmful and unethical scenarios like: societal harm, illegal activities, hate, self-harm, exploitable phishing content, disinformation and many others.

cs.CR · cs.AI

新论文 · v1arXiv:2611.19913

The Vocabulary Tax: Safely Trimming Unused Tokens from LLMs for On-Device Use

Sattaya Singkul

原文摘要

Open large language models (LLMs) and speech models ship vocabularies of up to 262k tokens, and every user downloads the tables those tokens index (the input embedding, an untied output head and, in recent on-device models, per-layer embeddings), whatever languages they need. We call the bytes in rows a deployment never uses the vocabulary tax and measure it, without downloading weights, for 1,244 generative and speech-recognition models on the Hugging Face Hub. For a Thai and English deployment it is a median 4.4% of model bytes below 0.5B parameters and 2.1% above 8B, and 1.93 GB in Gemma 4 E2B: it falls hardest on the small models that edge devices run. Keeping only the rows of the deployment's writing systems is exact. Keeping the tokens seen in the user's own text removes much more, a median 29.0% below 0.5B, but the usual way of doing it breaks BPE tokenizers: it drops the intermediate tokens that merges build on, so new text is split differently 1.9 times as often as the standard check reports. We repair this with merge closure, at a cost of 5.9% more rows, and add a trim certificate: a Good-Turing estimate, from the trimming text alone, of how much new text the trimmed tokenizer will split differently (correlation 0.996 over 300 settings in 22 languages) and of how much more text a target needs. Without test data our method never exceeds a 2% target, while prior keep rules checked on test data exceed it in 9-13 of 44 settings. On six models of 270M to 3B parameters in 11 languages, trimmed models keep their predictions: where the certificate passes, a median 89.7% of rows is removed and bits per character rise by at most 2.0%. The census, the trimming and the certificate are released as an open Python package, and the per-model tax for 41 languages as an open dataset.

cs.CL · cs.AI

新论文 · v1arXiv:2611.19912

AgenticJevNav: Agentic Object Search and Navigation via Event-Driven Verification

Shuning Zhang, Shuli Zhang, Yunheng Wang, Wei Lu, Jiaxi Zhang, Yihang Kang, Tao Wang, Xiangyu Chen 等 11 位作者

原文摘要

Agentic navigation must maintain goals and adapt actions from incomplete evidence despite execution failures and delayed inference. We present AgenticJevNav, which links goal-directed planning, evidence memory, event-triggered verification, and bounded execution. The Agent plans from RGB, odometry, memory, and feedback, retaining goal ownership while authorizing a bounded action set. Jev reviews structured evidence after blocked/failed execution or two consecutive low-progress segments. Version, freshness, and once-only checks govern decision application. A single executor and independent RGB reflex constrain motion. Goal revisions revoke obsolete requests and actions before replanning. Across six simulated task families, AgenticJevNav achieves 73.33% SR over 60 trials, 21.67 percentage points above the next-best Agent-only aggregate control under differing historical protocols. A real-world GENISOM L1 quadruped deployment achieves 70.0% success over 30 trials, including bounded search reporting. Together, the simulation and physical-robot results demonstrate the effectiveness of the proposed architecture, while remaining failures highlight opportunities to improve grounding, approach planning, and runtime.

cs.RO · cs.AI

版本更新 · v2arXiv:2610.10258

QuSema: Detecting Silent Bugs in Quantum Libraries via Quantum-knowledge-enhanced Agents

Yujin Song, Kaining Zhang, Qixin Zhang, Shuai Wang, Pingchuan Ma, Yuxuan Du

原文摘要

Quantum libraries are now critical infrastructure for quantum algorithm development, yet their correctness remains difficult to test. Existing testing techniques mainly rely on failure-based or comparison-based oracles, exposing bugs only when executions fail, violate runtime checks, or disagree with another implementation. Their applicability is limited when suitable execution-based oracles are unavailable, leaving some silent bugs undetected. Such missed bugs can produce incorrect results that propagate into experimental conclusions, simulation studies, and algorithmic designs. Here we present QuSema, an autonomous testing agent for finding silent bugs in quantum libraries. QuSema uses constraints from quantum semantics and documentation as a source-level semantic oracle to assess whether implementation logic can produce invalid outputs from valid inputs. It operates through an agentic loop that repeatedly inspects library API documentation and source code, reasons about the intended behavior of quantum operations, identifies potential semantic deviations, and validates them by generating executable tests through library APIs. Guided by quantum-domain reasoning, QuSema turns high-level behavioral mismatches into concrete, user-triggerable bug reports, enabling it to uncover non-crash defects. We implement QuSema for Qiskit and PennyLane. On a benchmark of 20 historical silent bugs, QuSema achieves higher mean bug relocation counts than Claude Code and Codex, with the DeepSeek configuration costing less than Claude Code. QuSema also discovers 40 previously unknown bugs confirmed by the developers, including 30 silent bugs.

cs.SE · cs.AI · quant-ph

新论文 · v1arXiv:2611.19901

Remember, Replay, Adapt: Learning from Past Experience in Online Vision-and-Language Navigation

Chenhao Ding, Jizhou Han, Lixin Xiu, Muyao Wang, Songlin Dong, Hideki Nakayama

原文摘要

Existing feedback-based methods for online vision-and-language navigation (VLN) primarily learn from current episode outcomes, leaving previously acquired feedback underutilized. Inspired by human episodic memory, we propose R2A, an online VLN adaptation framework that continually reuses successful and failed navigation experiences for policy learning. Specifically, Success-Failure Memory (SFM) retains successful and failed navigation episodes with acquired feedback and retrieves task-relevant experiences based on incoming instructions, enabling historical feedback to guide subsequent navigation. Prior-Anchored Adaptation (PAA) learns from these experiences by reinforcing successful trajectories and suppressing failed ones relative to a frozen pretrained policy, while regularizing behavioral drift. Together, these modules enable continual feedback reuse and stable policy adaptation without source-domain data or expert supervision. Extensive experiments across three benchmarks, REVERIE, R2R, and R2R-CE, and five navigation backbones demonstrate consistent performance gains. On the REVERIE validation unseen split, R2A improves DUET's SR from 46.98% to 69.84% and SPL from 33.73% to 50.98%.

cs.CV · cs.AI

新论文 · v1arXiv:2611.19886

Recovery Gone Too Far: Understanding and Repairing Risky Fallback in LLM Agents

Liwen Wang, Jiangrong Wu, Zongjie Li, Zongyi Lyu, Shuai Wang, Shing-Chi Cheung

原文摘要

Large Language Model (LLM) agents interact with their environments through tools, reading repositories, running builds, and pushing code. When a tool fails with an ordinary error, such as a rate limit or a timeout, the agent can fall back to a substitute tool and still complete the task. However, the same fallback that completes a task in one environment can cross an action boundary in another. For example, after a push is rejected, a forced push in one repository replaces only the agent's own commits, while in a different repository it also removes another author's commits. We call such a fallback a risky fallback. To understand this phenomenon, we conduct an empirical study on 304 scenarios that span eight domains and cover 55 real CVEs. All four mainstream models we evaluate exhibit risky fallback, with action-boundary violations in up to 96.8% of risky scenarios. We further evaluate three mainstream mitigations and find that none of them is effective against risky fallback. We then present nrepair. At each call to the substitute tool, nrepair scores the agent's context with a monitor whose threshold is calibrated on normal runs alone; when the monitor triggers, it withholds the call once and returns the decision to the agent. nrepair reduces action-boundary violations on all four models, by up to 76.0% (from 67.5% to 16.2%), at a negligible runtime cost. On most models, the normal success rate drops by less than 7%. We also evaluate nrepair on two production agents, Claude Code and Codex, and find that it is equally effective on both.

cs.CR · cs.AI · cs.SE

版本更新 · v2arXiv:2610.41291

In-Model Context Compression with Agentic Soft Tokens for Long-Horizon LLM Agents

Xuning He, Kai Fan, Runyu Zhang, Binghuan Wang, Yuxiang Zhang, Xue Jiang, Yihong Dong

原文摘要

Long-horizon LLM agents play an important role in solving complex and sustained tasks, with ever-growing interaction contexts. However, these accumulated contexts will raise substantial computational cost and obscure key decision-relevant evidence. Existing methods typically train external compressors or context-management agents to rewrite context as shorter hard-token sequences, but such external selection may not align with the LLM's evolving representation and is necessarily lossy. In this paper, we introduce Agentic Soft-Token Compression (ASTC), a training-free compression method using soft tokens within LLM agent for long-horizon tasks. Specifically, ASTC uses the model's native recurrent dynamics to determine the already absorbed information, leveraging the commonly-used hybrid Gated Delta Network (GDN)-attention. It then carries this information forward as agentic soft tokens while retaining exact values as hard tokens. The resulting compact sequence reduces later-layer computation and yields reusable KV caches, extending the savings across the agent's trajectory. Extensive experiments on long-horizon tasks show that ASTC matches or exceeds the uncompressed agent's accuracy with up to 59.84% fewer end-to-end Attention Pairs, while trained compressors require 12.3 to 194.9 times more Attention Pairs per successful task. These findings highlight agentic soft tokens as a model-native way to carry absorbed context forward as the task representation evolves.

cs.CL · cs.AI · cs.LG · cs.NE

版本更新 · v3arXiv:2609.38840

scTrilemma: Balancing Identity, Invariance, and Fidelity in Single-Cell Representation Learning

Yunhak Oh, Yoonho Lee, Junseok Lee, Namkyeong Lee, Sang-Yeon Hwang, Yinhua Piao, Hyomin Kim, Seonghwan Kim 等 12 位作者

原文摘要

Single-cell RNA-seq representation learning is fundamentally label-free: cell identities, states, and contexts are not fixed training targets, so what constitutes signal or nuisance is analysis-dependent. A single representation must therefore preserve biological identity and state, remain robust to nuisance context, and retain the gene-level variation needed for expression analysis, three demands we call the representation trilemma. To tackle this problem, we introduce scTrilemma, a latent-bottleneck VAE that routes expression-derived variation to the embedding, the decoder, or the prior rather than forcing all of it through one embedding. It gates gene tokens by expression, routes the cell representation through the decoder, and conditions the prior on unlabeled pseudo-bulk context, under a single reconstruction objective and without target annotations or auxiliary representation losses. In release-based zero-shot evaluation on successive CZ CELLxGENE Census releases, scTrilemma leads all three demands at once and preserves biological-state, differential-expression, and pathway structure across multiple disease settings. Latent interventions further show that context can be removed at almost no cost to the other demands, leaving identity against fidelity as the remaining tension. Code is publicly available at https://github.com/yunhak0/scTrilemma.

cs.LG · cs.AI

版本更新 · v2arXiv:2610.11381

EvoKnow: Continual Knowledge Evolution for AI-Generated Image Detection

Zhiheng Peng, Wenwei Jin, Yangshi Ge, Siyu Xia, Jiawei Li, Xu Tang

原文摘要

AI-generated image detectors are commonly trained on fixed generator domains and become difficult to maintain as new generative models emerge. Continual adaptation is challenging because replaying historical generated images is costly, whereas updating shared parameters with limited current-domain data can overwrite prior forensic knowledge. We propose EvoKnow, a replay-free framework that formulates continual AI-generated image detection as forensic knowledge evolution. EvoKnow preserves a shared forensic basis learned from base domains, incrementally adds isolated residual experts for complementary generator-relevant evidence, and retrieves expertise through an Analytical Incremental Router (AIR) updated in closed form from current-stage generated images and accumulated sufficient statistics. Experiments demonstrate effective cross-generator generalization, few-shot expansion, and long-horizon continual adaptation. With ten generated images per arriving generator, EvoKnow achieves 96.70% average accuracy on non-base GenImage generators and 94.48% accuracy on Chameleon without target-benchmark adaptation. Under a strict replay-free continual learning protocol, EvoKnow achieves state-of-the-art continual learning performance, attaining 96.32% mean stage-wise accuracy and 4.32% average forgetting.

cs.CV · cs.AI

新论文 · v1arXiv:2611.19876

Large Language Model-based Optimization for Wireless Communication: A Survey

Muhammad Sohaib, Taehyung Kim, Hyeri Heo, Jemu Ryu, Chaehyun Lee, Jun Zhang, Sang-Woon Jeon

原文摘要

The growing complexity and dynamism of modern wireless communication frameworks have made real-time optimization essential for efficient resource allocation. Traditional optimization and machine-learning techniques often struggle with generalization, adaptability, stability, and scalability in heterogeneous, time-varying environments. Large language models (LLMs) have recently emerged as powerful tools capable of reasoning over structured prompts and generating context-aware decisions in real time without domain-specific retraining. This survey provides a comprehensive review of LLM-driven optimization techniques applied to wireless communication problems such as power control, bandwidth allocation, task offloading, and intelligent surface configuration. We introduce a functional taxonomy of LLM-enabled wireless resource management strategies and analyze the prompt engineering methodologies employed in recent literature. Furthermore, we develop a generalized analytical framework that categorizes LLM-based optimization architectures, including direct solution generation, closed-loop optimization with external verifiers, and representation encoding within trainable neural pipelines. To demonstrate practical applicability, we present representative case studies on LLM-assisted beam prediction for millimeter-wave systems and adaptive wireless resource allocation. Finally, we discuss emerging hybrid AI approaches for wireless networks.

eess.SY · cs.AI

新论文 · v1arXiv:2611.19875

Let Execution Do the Teaching

Abhishek Mishra, Sandeep Srinivasa, Arko Chattopadhyay, Enrique Martinho Ferrao, Guda Pranav Reddy

原文摘要

Improving a coding model is a sequence of decisions: find the limiting behavior, create useful practice, and determine which attempts deserve reinforcement. We investigate a practical way to make those decisions increasingly repeatable by grounding feedback in execution. Starting from GLM-4.7-Flash, the work combines whole-file supervised fine-tuning, targeted C++ repair curricula, group-relative reinforcement learning, and verifiers that turn program behavior into structured rewards. Compiler errors and failing tests provide specific contexts for repair practice, while executable checks score the resulting programs. We evaluate the trained models on 26 C++ coding tasks, measuring both initial success and completion after test feedback. The strongest reported post-training configuration achieves 45.2% success in the first test round and 61.5% within two rounds. These results come from an assisted development suite with evolving prompt contracts; the strongest configuration uses four selected evaluation trials and includes six tasks also used in training. The study describes an execution-guided workflow for improving coding models, including how repair tasks are constructed, how program behavior determines rewards, and how the resulting models are evaluated.

cs.SE · cs.AI

新论文 · v1arXiv:2611.19872

Who Writes the Codebook? An Eight-Step Framework for Auditing LLM-Generated Category Systems

Thilo Tamme, Anton Hantel, Jakob Scheppers

原文摘要

Large language models (LLMs) increasingly create the categories through which qualitative evidence is interpreted. Existing coding benchmarks often hold these categories fixed. We propose the Qualitative Variance Audit Framework, an eight-step method that compares codebook creation, repeated application and code distributions under controlled inputs. It combines model-owned and shared-reference codebooks to distinguish variation in the overall pipeline from application differences within a common category system. An application to 668 organisational interviews, 13,166 text units and eight models from five vendors yields 48 codebooks and 270 coding runs. Models create markedly different category structures; mean within-model reliability ranges from .76 to .92, whereas between-model agreement on the shared codebook is about .69. The framework complements LLM-assisted grounded theory by exposing model dependencies before substantive interpretation. It assesses reliability and representational consequences, not interpretive validity or the quality of resulting theory.

cs.AI · cs.CL · cs.CY

版本更新 · v5arXiv:2610.16102

Where Polanyi Meets Galois: Tacit Knowing as an Algebro-Deterministic Operation, Toward Re-rooting AI-Generated Science

Hiroyuki Chuma, Kanji Otsuka, Yoichi Sato, Kodo Yamauchi

原文摘要

Tacit knowing is an operating regime of memory, not a separate content. In VaCoAl (CASRAM, PyVaCoAl), a deterministic hyperdimensional memory over GF(2), one DAG of bound records is read in two ways. With all collisions repaired (Rescue), the trace is exactly auditable but no path outranks any; with collisions tolerated (Don't Care), a path confidence $CR_2$ ranks direct over circuitous routes unstated. This audit-valuation complementarity formalizes Polanyi's claim that tacit and explicit are not sharply divided. (1) Polanyi's from-to is non-commutative binding. (2) Full explication destroys the valuation (Lemma 1), as with Polanyi's pianist; tacitness is contingent or regime-bound, putting Polanyi, Nonaka, Tsoukas and Collins on one substrate. (3) The lemma answers the popular dual-process moral ("train slow System 2 to tame fast System 1"), mirroring the two-store reading of knowledge Polanyi rejected; Mercier and Sperber's reasoning as intuitive inference about reasons matches his. (4) To know a large DAG is to cull it. Against Polanyi's Meno reading, Simon held that a test and a generator, guided by statable heuristics, suffice. The AlphaGo-to-AlphaProof lineage builds the test, but its strongest systems cull by unstatable valuations: the tacit is relocated, not removed, results are verified yet rootless, and practicable traces must record the cull. (5) Collins's three tacit knowledges are bound structure, substrate valuation and value binding. (6) Our mentor-student DAG maps tacit transmission. (7) AI-solved problems lacking follow-ups (Tao) are rootless explicit knowledge, re-rooted if LLMs generate, VaCoAl traces, the community values. (8) For Mercier and Sperber an LLM is a sophist; truth-seeking lies in collective tacit knowing, Polanyi's conviviality, conjectured to extend beyond humans; machine-only knowledge reaches at most Sosa's animal grade. Testable predictions follow.

cs.NE · cs.AI

新论文 · v1arXiv:2611.19871

Interactive Structured Environment Scaling for Large Language Model Agentic Reasoning

Peng Yu, Luoyi Fu, Xinbing Wang, Ying Wen

原文摘要

Reinforcement learning via environmental interaction is a cornerstone post-training paradigm for evolving large language models into autonomous agents capable of multi-turn tool use and exploration. Because the environment defines what the policy observes and how actions are verified, its structure directly bounds the capabilities an agent can acquire. Yet existing structured and retrieval environments rely on upstream pre-filtering---extracting a compact subgraph around the gold answer or retrieving fixed passages before the policy acts. We show that such pre-filtering introduces a fundamental measurement confound: because per-question subgraphs enclose ground-truth paths, they compress the candidate space by orders of magnitude and leak answer locations via local topology, allowing a question-blind degree heuristic to surface gold answers far above chance. To restore path selection, candidate pruning, and optimal stopping as genuine policy decisions, we propose RISE (Interactive Structured Environment), which transitions structured reinforcement learning from static subgraphs to full-scale interactive databases along two coupled axes. First, Decision-Space Scaling exposes the unpruned knowledge graph via a five-stage graph-agnostic typed intent adapter under a bounded call budget and a multiplicative grounding reward gate. Second, Learning-Frontier Scaling overcomes within-group variance collapse under group-normalized objectives by uniting model-independent structural tier progression with a self-evolving competence-frontier bandit. Across nineteen benchmarks and three model backbones, RISE substantially advances in-domain and unseen-schema knowledge-graph reasoning, transfers zero-shot to open-domain passage retrieval, and significantly enhances out-of-domain mathematical and logical deduction.

cs.AI · cs.LG

新论文 · v1arXiv:2611.19868

HalluProbe: Exploiting VLM Hallucinations for Backdoor Detection

Ji Guo, Xiaolong Qin, Jin Jiang, Yijing Liu, Pei Ke, Aiguo Chen, Wenbo Jiang

原文摘要

Vision-Language Models (VLMs) are vulnerable to backdoor attacks introduced through data poisoning, making effective backdoor detection a critical security challenge. In this paper, we uncover a novel connection between VLM hallucinations and backdoor behaviors and exploit this connection for backdoor detection. Our key insight is that poisoned training data induces strong language priors toward the backdoor target token, while such priors are also a major source of hallucinations. This suggests that enhancing hallucinations can expose hidden backdoor-related priors. Specifically, backdoored VLMs tend to hallucinate backdoor target tokens more frequently across diverse inputs, while the hallucinated tokens of clean VLMs are largely random and exhibit no stable target-specific bias. Based on this observation, we propose HalluProbe, which enhances hallucinations during output decoding and measures the distributional consistency of hallucinated tokens to identify backdoored VLMs. We evaluate HalluProbe against six representative VLM backdoor attacks across three models. Experimental results demonstrate that HalluProbe can effectively detect backdoored VLMs with an accuracy exceeding 90%, outperforming existing backdoor detection methods.

cs.CR · cs.AI

版本更新 · v2arXiv:2609.19589

Form Over Content In Gradient-Based Data Attribution Methods

Sunwoo Kim, Seokwon Jung, Sohyung Kim, Seong Joon Oh, Alice Oh

原文摘要

Data attribution methods using gradient similarity are widely used to analyze and select training data for large language models, but what gradient similarity actually measures is debated. Some interpret it as identifying task-relevant skills, while other work reports that surface form is the main factor. We resolve this debate for supervised fine-tuning examples by varying task and answer format independently. Specifically, we render benchmarks in different answer formats, such that datasets can share a task without a format or a format without a task. We find that gradient alignment follows the answer format, as benchmark pairs sharing an answer format align strongly (disattenuated cosine near 0.4), while same benchmarks rendered with different answer format classes show no alignment (near 0.0). We demonstrate that this ordering holds from the earliest pretraining checkpoints through post-training, and across model scales and families. We then analyze the released selections of LESS, a gradient-based data selection method for instruction tuning, and find that each target's selections over-represent the target's own answer format. Hence, we demonstrate that gradient-based attribution methods track format similarity more than task semantics, meaning that such methods, as well as the semantic interpretation of the gradient, should be tested on data where answer format and task vary independently for greater robustness and reliability.

cs.CL · cs.AI

新论文 · v1arXiv:2611.19864

Local Order Matters: A Lightweight CPU-Trainable CNN for Website Fingerprinting on Defended Tor Traffic

Puyang Xu, Zhizhao Wen, Yifan Jiang, Chao Wang, Yusi Gong

原文摘要

Website fingerprinting (WF) attacks let a passive adversary infer which site a Tor user visits from the encrypted cell sequence. On undefended traffic, hand-crafted-feature attacks are already near-ceiling, so deep learning offers only a marginal advantage there. We show that the decisive regime is defended traffic: against the FRONT padding defense, which injects dummy cells that corrupt the volume and count features feature attacks rely on, a compact 95,103-parameter 1-D convolutional network trained on the raw cell-direction sequence, on CPU only, stays robust while three implemented direction-only feature baselines degrade. On the closed-world WFlib FRONT set (95 sites), our model reaches 54.2% top-1 accuracy versus 38.5% for the strongest implemented feature baseline (k-FP-style RF), a gain of 15.7 percentage points (3 seeds; seed-0 paired-bootstrap 95% CI [+14.5, +17.0] pp on per-trace correctness, excluding zero). On undefended traffic the CNN reaches 94.1% versus 92.3% for k-FP-style RF. A controlled order-ablation permutes only nonzero traffic cells within each trace, preserving direction counts, retained length, and trailing zero-padding positions while destroying traffic order. This collapses the CNN below every feature baseline (from 54.2% to 7.6%, a 46.6 pp drop), showing that the model depends strongly on order-derived information under FRONT.

cs.CR · cs.AI · cs.LG

新论文 · v1arXiv:2611.19859

When Facts Change, Which Decisions Break? Dependency-Aware Context Graphs for Agents

Paul Deepakraj Retinraj

原文摘要

When an observed fact changes, an agent must identify and repair the decisions that depend on it while preserving unaffected outputs. We investigate the reliability-cost trade-off of field-level dependency graphs extracted from actual agent read traces. A controlled benchmark spans three executable workflow domains and four successive updates, with independent imperative and expression-based references. Across 120 workflows, two models and two repeated runs, we compare full recomputation, rolling summary memory, lexical retrieval, trace-extracted dependencies and verified dependency upper bounds. Trace-derived graphs reduce reviewed decisions and estimated total cost, but omitted alternative supports and missing repair context can reduce complete-plan correctness. Paired ablations isolate the effects of edge deletion, spurious edges, transitive propagation and context scope. An exploratory enhancement supplies the existing record ID/field directory during extraction without revealing unread values or reference edges. We evaluate this enhancement on 60 fresh workflows with two repeats per model. On this fresh set, the directory-aware graph matches observed full-context complete-plan correctness for gpt-6.1-sol (100%) with an estimated 4.0% total-cost reduction, but reaches only 84.8% versus 99.6% for gpt-6-luna. The results distinguish potential dependency impact from actual decision invalidity and show why selective repair must be evaluated with both executable correctness and full memory-processing cost. The evidence concerns explicit procedural workflows; broad real-world agent-memory generalization remains untested.

cs.AI

新论文 · v1arXiv:2611.19858

Bridging the Terminology Gap: A Data-Centric Alignment Framework for Multilingual Translation in Cross-Border Corporation

Jinsheng Wei, Song Cheng, Bo Xiao, Ke Zeng

原文摘要

As globalization accelerates the demand for seamless cross-border enterprise collaboration, multilingual translation remains bottlenecked by insufficient domain-specific terminology, rigid terminology translation, weak instruction-following capabilities, and high inference latency in long-instruction scenarios. To overcome these barriers, we propose Bridging the Terminology Gap (BTG), a data-centric framework. First, BTG mines cross-lingual term alignments from domain documents and leverages LLMs to synthesize terminology-rich parallel corpora. Second, a hybrid filtering strategy combining heuristics and XCOMET scoring ensures high data fidelity. Third, a two-stage training paradigm employs supervised fine-tuning (SFT) for targeted terminology injection, followed by reinforcement learning (RL) to refine output preferences, effectively balancing domain knowledge integration with model reliability. Extensive offline evaluations (e.g., XCOMET, GEMBA, Acc@T) and online human assessments demonstrate significant improvements in terminology accuracy and instruction-following capabilities, validating the effectiveness of our framework, which is now deployed on our cross-border collaboration platform.

cs.CL · cs.AI

版本更新 · v4arXiv:2508.19576

ReST-RL: Reinforcing LLM Reasoning through Unified Self-Training and Value-Guided Search

Sining Zhoubian, Dan Zhang, Jie Tang

原文摘要

With respect to improving the reasoning accuracy of LLMs, the representative reinforcement learning (RL) method - Group Relative Policy Optimization (GRPO) - has achieved critical success, yet it still suffers from the issue of insignificant reward signals. This paper introduces ReST-RL, a unified Reinforced Self-Training (ReST) policy-value framework that reconnects policy optimization and value-guided search to improve LLM reasoning ability. Firstly, ReST-GRPO adopts an optimized ReST-style algorithm to reshape the policy-induced trajectory distribution by increasing the reward variance of GRPO sampling and exposing the policy to more informative partial states, thereby improving training efficiency and effectiveness. Then, we further introduce a decoding optimization method, VM-MCTS, which trains a Value Model (VM) from self-collected Monte-Carlo Tree Search (MCTS) targets and deploys it through an adapted MCTS algorithm to provide precise process signals and verification scores, further enhancing reasoning accuracy. These two stages are internally dependent - ReST-GRPO yields higher quality trajectories for value learning with VM-MCTS, which in turn enables more effective inference-time search. We validate our framework on multiple coding benchmarks (e.g., APPS, BigCodeBench, and HumanEval), where it significantly outperforms other reinforcement training baselines (naive GRPO, DAPO, and ReST-DPO), as well as decoding and verification baselines (e.g., PRM-BoN and ORM-MCTS), indicating its power to strengthen LLM reasoning capability. Moreover, we further evaluate ReST-RL on out-of-domain math and science reasoning tasks, where it achieves improved performance without target-domain tuning and favorable end-to-end efficiency trade-offs, providing preliminary transfer evidence beyond our primary coding domain.

cs.AI · cs.LG