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

+0新增论文
+0其中 AI 相关
+0已有论文修订
+93新增 AI 导读

arXiv 最近完成扫描:2026-11-16 01:34 · 全站最近入库:2026-11-14 15:09各信源进度 →

统计口径与数据新鲜度

统计 2026-11-15 09:20 — 2026-11-16 09:20(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:2026-11-16 01:34最近发现新增版本:2026-11-14 15:09库内论文最新更新时间:2026-11-13 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。

23,348 已收录论文19,580 AI 相关6,729 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 23,348 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-16 01:33(北京时间)
版本更新 · v2arXiv:2611.11342

Agent Self-Improvement with Frozen Weights: The Added Value and Limits of Failure Traces

Arkid Mitra, Snehil Saluja

原文摘要

We evaluate whether an AI agent's harness, the notes around its model, can learn from the agent's failure traces beyond a matched search that never sees them. It can, but less than it looks. We tested an open 9-billion-parameter model with fixed weights; a larger teacher model wrote its notes. On generated customer-service decisions, where both note writers were told the target rules, notes based on the agent's mistakes improved accuracy by 4.0 percentage points over notes from the same process with the mistakes withheld. That process alone, trying and keeping better-scoring notes, reached about 60% of the total improvement. In full customer-service conversations the same notes showed no demonstrated improvement. Separately, a strict screen of single lessons admitted two customer-service lessons, still awaiting tests on new data, and one tool-use lesson, written without seeing the agent's mistakes, which showed no demonstrated gain on new tool-use tasks. A neutral note with no procedure scored higher than the learned lessons (exploratory). A blind audit judged 146 of 240 tool-use test items defective.

cs.LG · cs.AI · cs.CL

版本更新 · v2arXiv:2611.10508

Imperfectly Imperfect: Evaluating Graded Cognitive Diversity in LLM Persona Simulations

Sion Weatherhead, Ben Newell, Aaron Belbasis, Flora Salim

原文摘要

Large language model (LLM) agents are increasingly used as synthetic participants in social, behavioural, and decision-making research, yet it remains unclear whether they can express graded cognitive diversity as assigned when engaging in simulated behaviour. We introduce a cognition-suite evaluation that tests whether simulated populations reproduce human variation in reasoning, self-reported traits, their associations, and downstream behaviour. Agents conditioned on synthetic or human-derived profiles of cognitive reflection, Need for Cognition, and Big Five personality answer the same suite as a cohort of 473 adults: two newly assembled tests, a Cognitive Reflection Test (CRT) and a Behavioural Reflection Test (BRT) whose decision scenarios measure reflection in behaviour, alongside an established CRT and the two trait questionnaires. Across seven full-battery simulation runs of four models, agents recover Need for Cognition and Big Five trait distributions on self-report questionnaires, but fall short on simulating diversity of cognitive task performance and downstream behaviours. In the four Claude Sonnet 4.6 runs, every agent scores full marks on both CRTs, whatever reflection level its persona was assigned. In the other three runs agents make intuitive-lure errors (the tempting wrong answers CRT items invite), but 49 to 85% score full marks, against under 10% in the human sample. Claude Sonnet 5 comes closest to people's item-by-item pattern of lures; the other two runs give fewer lures and vary less between items. In the BRT scenarios, every run over-produces 8 to 11 of 15 coded behaviours by at least 20 percentage points, and people's links between reflection and behaviour carry over only in part. Useful synthetic populations must preserve not only trait descriptors but the structured variation in reasoning, error type, and behaviour those descriptors are meant to organise.

cs.CL · cs.AI · cs.CY

新论文 · v1arXiv:2611.14901

Understanding the Effects of Multi-Agent Deliberation for Human-AI Decision-Making

Harry Yizhou Tian, Ming Yin

原文摘要

Multi-agent deliberation (MAD), where multiple LLMs engage in structured discussion, improves AI performance on many benchmarks, but whether it helps humans make better decisions remains untested. We conduct two user studies on LLM-response-verification and fact-checking. We first compare MAD against common AI decision-support methods that isolate its two defining dimensions --- multiple AIs and multi-round interaction. We find MAD outperformed single-round AI assistance on both tasks; multi-round interaction was the main driver of this benefit, while a second AI added value when the single AI had room to improve. Then, we study two design choices in MAD --- AI persona and human engagement level --- finding a devil's advocate persona to lower accuracy and user experience, while humans' active participation in MAD improved accuracy in one task at the cost of frustration. We conclude by discussing implications for when, why, and how multi-agent deliberation can support human decision-making.

cs.HC · cs.AI

新论文 · v1arXiv:2611.14900

Nested Ensemble Learning with Topological Data Analysis for Graph Classification and Regression

Innocent Abaa, Umar Islambekov

原文摘要

We propose a nested ensemble learning framework that utilizes Topological Data Analysis (TDA) to extract and integrate topological features from graph data, with the goal of improving performance on classification and regression tasks. Our approach computes persistence diagrams (PDs) using lower-star filtrations induced by three filter functions: closeness, betweenness, and degree2 centrality. To overcome the limitation of relying on a single filter, these PDs are integrated through a data-driven, three-level architecture. At Level-0, diverse base models are independently trained on the topological features extracted for each filter function. At Level-1, a meta-learner combines the predictions of these base models for each filter to form filter-specific ensembles. Finally, at Level-2, a meta-learner integrates the outputs of these filter-specific ensembles to produce the final prediction. We evaluate our method on both simulated and real-world graph datasets. Experimental results demonstrate that our framework consistently outperforms base models and standard stacking methods, achieving higher classification accuracy and lower regression error. It also surpasses existing state-of-the-art approaches, ranking among the top three models across all benchmarks.

stat.ML · cs.LG · stat.AP

新论文 · v1arXiv:2611.14898

From Pixels to Prompts: Physics-Guided Language Enhancement for AI-Generated Image Detection

Mei Qiu

原文摘要

AI-generated content (AIGC) has substantially narrowed the visual gap between natural and synthetic images, while existing detectors often rely on generator-specific artifacts and exhibit limited generalization to unseen models. This motivates a fundamental question: which intrinsic physical image characteristics provide stable, discriminative cues for distinguishing natural from AI-generated images? We investigate this by systematically identifying robust physical features and incorporating them into a vision-language representation without external textual prompts at inference. We first analyze 15 handcrafted physical descriptors across images from over 20 generative models. Based on cross-generator stability and real-fake discriminability, we propose a Feature Stability-Discriminability Validation Assessment (FSDVA) to select a compact set of reliable features. Five core features, including Laplacian variance, Sobel statistics, and residual-noise variance, are then used to construct structured textual descriptions encoding physical evidence for CLIP-based learning. We integrate these physics-enriched descriptions with image representations via a native physics-guided multimodal framework, adapting CLIP ViT-L 14 with LoRA. At inference, the model requires only the input image, with no external captions or manual prompts. On GenImage, our method achieves 99.0 percent accuracy on SDv1.4 and 96.7 percent on Wukong, with a mean of 97.23 percent across eight generative models versus 95.8 percent for C2pCLIP under the same setting. On UniversalFakeDetect, it maintains strong, stable generalization across diverse and unseen generation settings. These results show that carefully selected physical image statistics provide complementary forensic evidence to pretrained vision-language representations, enabling interpretable, more generator-agnostic AIGC detection.

cs.CV

新论文 · v1arXiv:2611.14897

Social Engineering Attack Detection in Corporate Communications

Hanzla Hamid, Pappu Jha, Nick Rahimi

原文摘要

Social Engineering attacks continue to pose a significant risk to the corporate world with a growing level of sophistication. It is difficult to distinguish between authentic and phishing communication via email, chat, and phone. They pose a risk of exposing trade secrets and confidential communications and can create a chain of attack from lower-level employees up to executives. Therefore, we study this growing threat using real-world datasets to build baseline machine-learning models. Our TF-IDF models classify emails as phishing or safe, achieving accuracy above 90%. We build on this by exploring whether psychological manipulation and persuasion indicators can be measured to play a significant role in detecting phishing. Instead of relying solely on surface-level lexical features, we derive interpretable lexicon-based indicators of urgency, trust, greed, and other persuasion cues from the combined email dataset. By evaluating these manipulation-oriented features on their own, we measure how much signal they carry (ROC-AUC of 0.74 on a held-out test split) and test whether they transfer to held-out email sources.

cs.LG · cs.CR

新论文 · v1arXiv:2611.14895

HeadIO: Body-Frame Heading Decomposition for Neural Inertial Odometry on Head-Mounted Devices

Taeyoung Yeon, Yiquan Li, Chenfeng Gao, Vasco Xu, Karan Ahuja

原文摘要

Inertial odometry (IO) using only Inertial Measurement Units (IMUs) enables camera-free head tracking in head-mounted devices such as VR headsets, AR glasses, and earbuds. Prior IO methods operate in a gravity-aligned world frame, which is not yaw-equivariant. Yaw augmentation improves robustness but does not guarantee equivariance; learned equivariant layers achieve it, but EqNIO's full pipeline costs nearly 3$\times$ the FLOPs of ours. While raw body-frame IMU inputs are inherently yaw-invariant, head tilt redistributes both gravity and motion signals across device axes, making the representation tilt-variant. We propose HeadIO, which combines a body-frame heading decomposition with a bidirectional Mamba backbone for head-mounted inertial odometry. Our decomposition achieves tilt invariance by separating gravity-aligned and horizontal components using only the tilt component of device orientation. We evaluate on four head-mounted datasets and find that HeadIO improves RTE-5s by 8.1% (0.260 m $\to$ 0.239 m) over world-frame methods with learned equivariant layers on Nymeria, the largest benchmark dataset, at nearly 3$\times$ lower FLOPs and 8$\times$ fewer parameters. HeadIO achieves state-of-the-art performance on our primary metric, RTE-5s, on all Aria-glasses datasets, which natively provide body-frame semantics for our decomposition method. We further evaluate the IMU-only deployment scenario on Nymeria, where HeadIO remains state-of-the-art. We show that our method relies mainly on tilt accuracy, which can be obtained from the gravity signal; for yaw, a single-coefficient gyroscope correction removes 11.6% of the yaw error and improves RTE-5s by 2.5%, and sparse external yaw corrections (for example, one every 30 s) can improve RTE-5s by 12.0%. Our decomposition is lightweight, composes with any backbone, and is suitable for real-time deployment on head-mounted devices.

cs.CV · cs.RO

新论文 · v1arXiv:2611.14894

Machine Learning Moment Closure for the Boltzmann Transport Equation: Enforcing Hyperbolicity

Bongseok Kim, Jiahao Zhang, Guang Lin

原文摘要

The Boltzmann transport equation describes particle transport across collisional regimes, but resolving its velocity dependence is expensive. Moment methods reduce this cost by evolving finitely many velocity moments, requiring a closure for unresolved higher-order moments. Data-driven closures can represent these moments accurately, yet without structural constraints they may produce complex characteristic speeds and unstable simulations. Learned Boltzmann closures with guaranteed hyperbolicity have so far been formulated in one spatial dimension; in multiple dimensions, hyperbolicity must hold in every direction simultaneously. We develop, to the best of our knowledge, the first data-driven closure for multidimensional Hermite moment systems of the linear Boltzmann equation that is hyperbolic by construction. The network learns a symmetric positive-definite symmetrizer shared by all spatial directions and restricted to a subspace that preserves the known lower-order transport equations; the unresolved rows follow from a symmetric completion. Every directional Jacobian thus has real eigenvalues and a complete set of eigenvectors for every network output and at any truncation order. The closure is trained only on spatial derivatives of unresolved moments from kinetic reference solutions, with no solver in the training loop, and evolved by a partially conservative discontinuous Galerkin method. In two-dimensional experiments in the transition and scattering-dominated regimes, a single learned third-order system with ten moments remains stable in all cases. Averaged over time, it is more accurate than the classical P7 and P15 truncations with 36 and 136 moments in every experiment, and its density error is about two orders of magnitude smaller than that of the P3 truncation with the same moments.

cs.LG · math.NA

版本更新 · v2arXiv:2603.01478

Contract-Theoretic Semantic Abstraction for Covert Communication in LAENets

Yansheng Liu, Jinbo Wen, Kun Zhu, Hui Li, Yang Zhang, Jiawen Kang

原文摘要

In Low-Altitude Economy Networks (LAENets), Unmanned Aerial Vehicles (UAVs) transmit task-related information to the Base Station (BS) under stringent low-probability-of-detection constraints. These constraints limit transmission resources, hindering reliable information delivery through conventional bit-level communication. Semantic Communication (SemCom) addresses this challenge by prioritizing task-relevant information over bit-level accuracy. However, semantic abstraction must balance information loss and redundancy to preserve task reliability and transmission efficiency. Moreover, the BS cannot directly observe or reliably verify UAVs' semantic processing capabilities and abstraction-dependent transmission costs, leading to information asymmetry. We propose an incentive-aware semantic abstraction framework for covert communications in LAENets. Specifically, we adjust UAV-side semantic abstraction levels to provide differentiated covert semantic services under fixed-dimensional transmission payloads. To address information asymmetry, we develop a contract-theoretic model and adopt Prospect Theory (PT) to capture risk-sensitive semantic abstraction decisions under uncertainty. We further design a Regularized Diffusion-based Soft Actor-Critic (RDSAC) algorithm for optimal contract design under PT, incorporating diffusion entropy and action entropy regularization. Numerical results validate the proposed framework and show that RDSAC improves average rewards by 6.77% over SAC and 31.78% over proximal policy optimization.

cs.NI

新论文 · v1arXiv:2611.14892

Almost Optimal Lower Bounds under Almost ETH for Approximating Euclidean k-means

Karthik C. S., Dor Minzer

原文摘要

In the Euclidean $k$-means problem, we are given a set of $n$ points in $\mathbb{R}^d$, and the goal is to find $k$ representative points in space so as to minimize the sum of the squared distances from each point to its closest representative. Algorithmically, de la Vega, Karpinski, Kenyon, and Rabani [STOC'03] and Kumar, Sabharwal, and Sen [JACM'10] obtain a $(1+\varepsilon)$-approximation for high-dimensional Euclidean $k$-means in time $2^{(k/\varepsilon)^{O(1)}} \cdot dn^{O(1)}$, which was subsequently improved to $2^{\tilde{O}(k/\varepsilon)} + O(ndk)$ by Feldman, Monemizadeh, and Sohler [SoCG'07]. On the lower-bound side, a recent work of Cohen-Addad, Karthik, Saulpic, and Schwiegelshohn [SoCG'26] proposed the Exponential Time for Expanders Hypothesis (XXH), which asserts exponential-time hardness for $(1+\varepsilon)$-approximating Vertex Cover on vertex expanders, and used it to prove nearly matching running time lower bound whenever $k=\tildeω(1/\varepsilon)$. We observe that XXH does not hold in the parameter regime needed for the above result, and show how to repair their reduction under a more standard assumption. More precisely, we show that assuming the Almost Exponential Time Hypothesis for 3-SAT (Almost ETH), for every constant $β\in(0,1)$, Euclidean $k$-means in $\mathbb{R}^d$ has no $(1+\varepsilon)$-approximation algorithm running in time $2^{(k/\varepsilon)^{1-β}}\text{poly}(n,d)$, whenever $k=\tildeω(1/\varepsilon)$. By modifying the construction, we also obtain a similar result for approximating the Euclidean $k$-median problem.

cs.CG · cs.CC · cs.DS · stat.ML

版本更新 · v5arXiv:2603.28590

MonitorBench: A Comprehensive Benchmark for Chain-of-Thought Monitorability in Large Language Models

Han Wang, Yifan Sun, Brian Ko, Mann Talati, Jiawen Gong, Zimeng Li, Naicheng Yu, Xucheng Yu 等 11 位作者

原文摘要

Large language models (LLMs) can generate chains of thought (CoTs) that are not always causally responsible for their final outputs. When such a mismatch occurs, the CoT no longer faithfully reflects the actual reasons (i.e., decision-critical factors) driving the model's behavior, leading to the reduced CoT monitorability problem. This limits the use of CoTs for reliable oversight. However, a comprehensive and fully open-source benchmark for thoroughly evaluating CoT monitorability remains lacking. To address this gap, we propose MonitorBench, a systematic benchmark for evaluating CoT monitorability in LLMs. MonitorBench provides: (1) a diverse set of 1,540 test instances with carefully designed decision-critical factors across 19 tasks spanning 7 categories to characterize when CoTs can be used to monitor the factors driving LLM behavior; and (2) two prompting stress-test settings to quantify the extent to which CoT monitorability can be degraded. Extensive experiments show that CoT monitorability is a conditional property affected by the evaluated LLM, monitor LLM, and task characteristics. Across these factors, monitorability is higher when decision-critical factors shape the intermediate reasoning process, rather than merely influencing the final answer. Under stress-test prompting, most evaluated LLMs can intentionally reduce monitorability, mainly on tasks where decision-critical factors are not structurally required by the reasoning process. Overall, MonitorBench provides a basis for further research on AI control, reasoning faithfulness, stress-test monitorability, and monitoring scaffords. The code is available at https://github.com/ASTRAL-Group/MonitorBench.

cs.AI

新论文 · v1arXiv:2611.14889

Can an AI Agent Be Epistemically Akratic? Higher-Order Evidence & the Limits of Self-Correction

Eno Agolli

原文摘要

When we judge that our reasoning was defective, it seems plausible that we should abandon the conclusion it produced. This thought has special force for AI agents built to assess and correct their own work. Yet obeying such judgments can make an agent worse at distinguishing sound reasoning from defective reasoning. An AI agent may rely on information about its reasoning when deciding which conclusions to keep, while the part that judges whether those conclusions are justified cannot use that information. The agent can then exercise an epistemic competence by retaining conclusions it judges unjustified, even when its self-assessments are generally reliable. Nor is the difficulty resolved by training the agent to make its conclusions and assessments agree: agreement can be achieved by degrading either one. Crucially, I argue that a negative assessment can remain valuable by identifying conclusions requiring further scrutiny, without having authority to eliminate them. The result challenges a central assumption about AI self-correction: that bringing an agent's beliefs into agreement with its own self-assessment about their justification makes it better at correcting its mistakes.

cs.AI

新论文 · v1arXiv:2611.14887

Characterizing Local Breadth in Language Generation

Jon Kleinberg, Fan Wei

原文摘要

Large language models sharpen the validity--breadth trade-off: a generator should avoid hallucinations without collapsing onto a narrow set of valid outputs. We study generation from positive examples of an unknown target in the Kleinberg--Mullainathan model, with arbitrary, possibly adversarial order and no distributional or computational assumptions. Breadth is measured by lower Banach density, which rules out arbitrarily large sparse holes, unlike prefix-based lower asymptotic density. 1. Exact optima in one dimension: For each target $K$ in a countable family, we characterize its optimal lower Banach density $b(K)$ via equivalent combinatorial and topological formulations. One eventually valid generator attains all $b(K)$ simultaneously. Every $α\in[0,1/2]$ occurs as $b(K)$ in some countable family; finite Cantor--Bendixson point rank guarantees $b(K)=1/2$, even without a uniform rank bound. 2. Unavoidable trade-offs in higher dimensions: For $d\ge2$, there need not be a single best generator: optimizing density for one target may worsen the guarantee for another. An explicit family quantifies this intrinsically multi-objective trade-off. 3. Complete characterization of density profiles: We characterize the full region of simultaneously achievable guarantees, with equivalent combinatorial and topological formulations, after excluding geometrically degenerate rectangles. Every prescribed product $\prod_i[0,α_i]$, $α_i\in[0,1/2]$ can occur as an exact achievable region for some language collection. 4. Near-half guarantees and the sharp endpoint: Under a near-logarithmic nondegeneracy condition, one generator achieves filtered density at least $1/2-\varepsilon$ at all finite-point-rank targets, for every $\varepsilon>0$, improving the previous polylogarithmic threshold. The exact $1/2$ endpoint can fail even for a rank-zero singleton family.

cs.DS · cs.FL · cs.LG · stat.ML

版本更新 · v2arXiv:2611.07487

When Does Blockwise Additivity Break? Diagnosing and Repairing the Additive Assumption in Distillation-Based Architecture Search

Yuefeng Xu, Jun Yu

原文摘要

Blockwise distillation methods for architecture search, such as DNA, DONNA, LANA and Puzzle, score each candidate replacement block in isolation and combine the scores additively. We ask at which budgets this additivity assumption fails, whether fine-tuning repairs the failure, what diagnoses it, and how it can be repaired cheaply. On DeiT-S, DeiT-B, Swin-T and Qwen2.5-0.5B with seven latency budgets each, an additive integer program trails a pairwise interaction-aware program after fine-tuning by at most 1.6 top-1 points on ImageNet-1k at the mildest budgets, but by up to 17.5 points, and a factor of 1.9 in perplexity, at aggressive ones. After 10,000 fine-tuning steps 3.3-9.9 points remain on DeiT, whereas on Swin-T the gap shrinks to at most about 1.8 points within 500 steps. In a learning-rate sweep with the standard 2,000 steps, 6.0-10.7 points remain on DeiT when each method's rate is chosen by its evaluation-set accuracy. An intervention that keeps the block library and only forbids more than two removals in any three adjacent layers shrinks the DeiT gap from 5.8-17.5 to 0.7-5.2 points, mostly by improving the additive solution: adjacent removals are a major failure pattern. Splitting the additivity error into a metric and a functional part, relative to additive logit changes, indicates where a cheap logit-sum predictor can be trusted. A hybrid search that ranks moves with this predictor and measures only the best needs no pair measurements and, after Holm adjustment, is significantly better than the pairwise program in 10 of 14 ImageNet-1k settings, with no statistically significant losses. An ablation shows that its ranking helps except at the most aggressive DeiT-B budget, where the predictor misleads and random ranking or ten-times-costlier evolution does better.

cs.NI · cs.NE

新论文 · v1arXiv:2611.14883

Random feature policy iteration for inverse mean field games with complex geometry and high dimensions

Xiangjian Zhang, Shanyin Tong, Xiaofeng Xu

原文摘要

We propose a random feature policy iteration method for reconstructing the obstacle function in a mean-field game (MFG) from observations of its value function, leveraging the method's mesh-free nature and flexibility to handle complex geometries and high-dimensional problems. The method builds on a policy iteration framework for inverse MFGs, which decouples the coupled nonlinear forward-backward PDE system into a sequence of linear PDE and inverse problem solves. The fixed random feature representations preserve this linear structure, allowing the resulting subproblems to be solved by minimizing residuals at sampled collocation points through linear least-squares problems. The mesh-free nature of the method further extends the applicability of the policy iteration framework to inverse MFGs with complex geometry and in high dimensions. Another key advantage is its flexibility in using different spatial approximations for different variables: the obstacle function, value function, and population density can be represented in separate feature spaces, accommodating obstacles that are substantially less regular than the corresponding PDE solutions. Moreover, boundary conditions can be enforced through collocation on irregular domains without constructing boundary-fitted meshes. Numerical experiments involving smooth and discontinuous obstacles, a domain with holes, and higher-dimensional problems demonstrate the accuracy and flexibility of the proposed method.

math.NA · cs.GT · math.OC

新论文 · v1arXiv:2611.14881

MLIRBench - A Benchmark and Dataset for Compiler and Hardware-Aware Reasoning

Pranava Sai Maganti, Meghasyam Peddireddy, Muzakker Hossain, Kaining Gao, Farhan Ahmed, Ali Jannesari

原文摘要

In recent years, large language models (LLMs) have become increasingly popular in software engineering and have performed well on different programming tasks. However, it is still difficult to know whether these models truly understand program behavior or whether they mainly rely on patterns learned from their training data. This distinction is particularly important for intermediate representations (IRs), where operations, dataflow, and execution semantics are explicitly represented. In this paper, we introduce MLIRBench, a benchmark for evaluating how LLMs reason about MLIR programs. MLIRBench uses a multi-agent pipeline to generate program mutations, validate their behavior through execution, and construct reasoning tasks from torch and linalg program pairs. The dataset contains 1,768 validated pairs, including 888 linalg pairs and 880 torch pairs, labeled as EQUIV, DIVERGE, or ERROR. We evaluated nine models, including five open-weight and four proprietary models. Gemini 3.1 Pro achieved the highest hard-label accuracy of 92.65% across all 1,768 pairs, counting unsuccessful predictions as incorrect. Across its 1,768 scored predictions, it achieved an aggregate score of 74.18 and a judge-assigned result-behavior score of 5.04/10. These results show that strong semantic classification does not necessarily imply accurate prediction of detailed execution behavior, highlighting a gap in current LLM reasoning over compiler IRs. By exposing this gap through execution-validated program pairs, MLIRBench provides a reproducible foundation for studying semantic reasoning over compiler IRs and related tasks such as program repair and semantic classification.

cs.PL · cs.AI · cs.LG

新论文 · v1arXiv:2611.14880

FOE-Dreamer: Deployment-Efficient Learning of Cyber Defense Policies in Operational Networks

Yinuo Du, Christopher Kiekintveld

原文摘要

Autonomous cyber defense is widely studied with deep reinforcement learning, but most work trains policies inside simulators because learning requires many trial-and-error interactions. Simulator-based training carries a fidelity gap: trained policies face authentication delays, partial action failures, service dependencies, and user traffic patterns that the simulator did not represent, and degrade on transfer. Training directly in the network being defended removes the gap but makes every interaction slow and costly, so sample efficiency becomes the binding constraint. Model-based reinforcement learning has reduced sample requirements on games and physical robots by orders of magnitude by planning inside a learned world model, but it has not been applied to defenders trained in operational networks. We present FOE-Dreamer (Factored Opponent-Enhanced Dreamer), a model-based reinforcement learning framework that trains a defender online and directly in a network on cloud infrastructure, without a simulator stage. FOE-Dreamer separates three sources of variation in the latent state: a controllable component driven by defender actions, an exogenous component driven by benign user activity, and an opponent embedding inferred from defender-observable signals alone. Separating the components reduces the variance in policy gradients during planning in imagination. Against two scripted attacker profiles with user activity emulated by the GHOSTS framework on an 8-host OpenStack testbed, FOE-Dreamer trains within a 3-day budget on a single GPU and achieves about 2$\times$ better episode loss than Rainbow and IQN baselines under matched compute. Both the factoring and the opponent model contribute to the gain, and the learned latents are consistent with the intended separation.

cs.LG · cs.CR

版本更新 · v3arXiv:2208.04433

Peer Prediction for Learning Agents

Shi Feng, Fang-Yi Yu, Yiling Chen

原文摘要

Peer prediction refers to a collection of mechanisms for eliciting information from human agents when direct verification of the obtained information is unavailable. They are designed to have a game-theoretic equilibrium where everyone reveals their private information truthfully. This result holds under the assumption that agents are Bayesian and they each adopt a fixed strategy across all tasks. Human agents however are observed in many domains to exhibit learning behavior in sequential settings. In this paper, we explore the dynamics of sequential peer prediction mechanisms when participants are learning agents. We first show that the notion of no regret alone for the agents' learning algorithms cannot guarantee convergence to the truthful strategy. We then focus on a family of learning algorithms where strategy updates only depend on agents' cumulative rewards and prove that agents' strategies in the popular Correlated Agreement (CA) mechanism converge to truthful reporting when they use algorithms from this family satisfying an additional condition on full exploitation. This family of algorithms is not necessarily no-regret, but includes fixed-parameter versions of familiar online learning algorithms (e.g multiplicative weight update and Follow the Perturbed Leader) as special cases. Simulation of several algorithms in this family as well as the $ε$-greedy algorithm, which is outside of this family, illustrates their finite-horizon behavior in the CA mechanism.

cs.GT

新论文 · v1arXiv:2611.14879

Typify: A Tool for Usage-driven Type Inference in Python

Hassan Aman, Ali Aman, Muhammad Asaduzzaman, Shaowei Wang, Jessica Chen

原文摘要

Python's dynamic type system, while expressive and flexible, creates significant challenges for static analysis tools in unannotated or partially annotated codebases. Existing approaches either require pre-existing type annotations, or depend on deep learning models that demand large training corpora and significant computational resources. We present Typify, a lightweight, usage-driven static type inference engine that infers precise type information across entire Python projects without annotations or training data. Typify integrates symbolic execution, fixpoint analysis, and context-matching retrieval to propagate types across modules, functions, and recursive call structures. The tool is available as both a command-line interface (CLI) for batch inference, evaluation and downstream integration with other tools, and a Visual Studio Code (VS Code) extension for interactive type assistance. Evaluation on the ManyTypes4Py and Typilus datasets shows that Typify outperforms all static baselines and closely matches state-of-the-art hybrid deep learning approaches, while running approximately 10x faster.

cs.SE

新论文 · v1arXiv:2611.14878

Hear What You Eat: In-Ear Passive Acoustic Sensing for Multi-task Dietary Monitoring

Feiyu Han, Mingtian Zhu, Dawei Yan, Weiwei Jiang, Yuanhao Feng, Yanfei Zhang, Panlong Yang

原文摘要

The increasing obesity health concerns are driving the advancement of continuous dietary monitoring. Yet, existing single-sensor solutions still face challenges of coarse-grained sensing resolution, while multi-sensor fusion strategies encounter significant engineering complexity or hardware requirements. In our work, we present a single modality-based multi-dimensional dietary monitoring and logging solution, named EARDIET, which leverages inward-facing microphones on earables to capture in-ear chewing sounds for fine-grained diet analysis. With the texture audio profiles derivation, the pipeline consisting of signal processing and lightweight deep learning, is carefully designed to achieve eating episode detection, food texture sensing, and food type recognition simultaneously. Comprehensive experiments involving 32 participants with diverse BMI profiles are conducted to validate the effectiveness of multi-dimensional dietary monitoring. Compared with the state-of-the-art solutions, EARDIET can derive multi-dimensional dietary information with only a single modality, outperforming by about 10% in terms of eating episode detection. Furthermore, through real-world clinical validation, our solution can achieve masticatory disorder identification with an averaged F1-score of 85.97%, demonstrating potential for healthcare in the future.

cs.HC

新论文 · v1arXiv:2611.14877

Message-Passing-Based Bayesian Joint Channel Estimation and Data Detection for Pilot-Guard-Free AFDM

Yuan Zhengdao, Guo Qinghua, Jin Ming, Liu Zilong, Xiao Pei

原文摘要

Existing channel estimation methods for affine frequency division multiplexing (AFDM) over doubly dispersive channels commonly isolate pilots and data with pilot guard symbols, resulting in substantial transmission overhead, while fractional Doppler introduces off-grid mismatch and complicates accurate channel estimation. This paper develops a pilot-guard-free AFDM transmission and reception scheme with a Bayesian receiver for joint channel estimation and data symbol detection. Uniformly interleaved pilots are embedded among data symbols without dedicated guard symbols, and the resulting pilot-data coupling is explicitly incorporated into the joint inference problem rather than suppressed through guard-based isolation.

cs.IT

版本更新 · v2arXiv:2409.02810

A hybrid FEM-PINN method for time-dependent partial differential equations

Xiaodong Feng, Haojiong Shangguan, Tao Tang, Xiaoliang Wan

原文摘要

In this work, we present a hybrid numerical method for solving evolution partial differential equations (PDEs) by merging the time finite element method with deep neural networks. In contrast to the conventional deep learning-based formulation where the neural network is defined on a spatiotemporal domain, our methodology utilizes finite element basis functions in the time direction where the space-dependent coefficients are defined as the output of a neural network. We then apply the Galerkin or collocation projection in the time direction to obtain a system of PDEs for the space-dependent coefficients which is approximated in the framework of PINN. The advantages of such a hybrid formulation are twofold: statistical errors are avoided for the integral in the time direction, and the neural network's output can be regarded as a set of reduced spatial basis functions. To further alleviate the difficulties from high dimensionality and low regularity, we have developed an adaptive sampling strategy that refines the training set. More specifically, we use an explicit density model to approximate the distribution induced by the PDE residual and then augment the training set with new time-dependent random samples given by the learned density model. The effectiveness and efficiency of our proposed method have been demonstrated through a series of numerical experiments.

math.NA · cs.AI

新论文 · v1arXiv:2611.14871

Women's Genetic Health as Family Cooperative Work: Empirical Findings and a Case for Participatory AI

Ziqi Yang, Yunan Chen, Daniel A. Epstein

原文摘要

Genetics plays a significant role in women's health, with family members often experiencing shared symptoms or hereditary risks across generations. Yet health technologies, including emerging health AI tools, have supported women's health as an individual health practice. Drawing on a survey of 249 U.S. adult women, we characterize women's genetic health information work as cooperative work distributed across intergenerational family networks, identify who does this work and why, and document the barriers that fragment it. We find that generative AI use is a strong predictor of genetic health perception, but current tools focus on individuals rather than supporting collective coordination. We argue that designing AI for this domain requires participatory approaches that engage diverse women, healthcare professionals, and legal and community stakeholders, and we bring open design questions to the workshop.

cs.HC

新论文 · v1arXiv:2611.14870

Synthetic Data Augmentation for Fairness under Data Imbalance: Theory and Adaptive Allocation

Zhengchi Ma, Kaifeng Zhang, Anru R. Zhang

原文摘要

Data imbalance across sensitive groups and class labels is common in classification problems and can contribute to substantial fairness disparities. We study how such imbalance affects Equal Opportunity and how synthetic data augmentation can mitigate the resulting unfairness. We first identify two mechanisms, prevalence imbalance and group-size imbalance, through which changes in the training distribution shift the learned decision boundary and induce disparities in group true positive rates. Motivated by this analysis, we study exact full-balance augmentation, which equalizes all group-label cells. We establish conditions under which full balancing strictly improves Equal Opportunity together with a prediction-error comparison under the oracle balanced distribution, and show that the Equal Opportunity gap of the synthetic-data classifier converges to that of the oracle balanced classifier as the sample size grows and the generator error vanishes. The finite-sample convergence rate agrees, up to logarithmic factors and generator-approximation error, with the information-theoretic lower-bound rate under the corresponding regularity regimes. Since exact balancing need not provide the best fairness-accuracy trade-off in practice, we further propose Accuracy-Constrained Adaptive Synthetic Augmentation (ACASA), which learns both where and how much to augment by minimizing estimated fairness disparity subject to a constraint on predictive accuracy. We provide a finite-sample guarantee comparing the selected augmentation allocation with the population-optimal accuracy-feasible allocation. Simulation studies and real data application support the theoretical findings and demonstrate that synthetic augmentation can substantially reduce fairness disparities while achieving favorable empirical fairness-accuracy trade-offs.

stat.ML · cs.LG · stat.ME

版本更新 · v3arXiv:2603.00763

Analyzing and Improving Training-Free Fast Sampling of Text-to-Image Diffusion Models

Zhenyu Zhou, Defang Chen, Siwei Lyu, Chun Chen, Can Wang

原文摘要

Text-to-image diffusion models have achieved unprecedented success but still struggle to produce high-quality results under limited sampling budgets. Existing training-free sampling acceleration methods are typically developed independently, leaving the overall performance and compatibility among these methods unexplored. In this paper, we bridge this gap by systematically elucidating the design space, and our comprehensive experiments identify the sampling time schedule as the most pivotal factor. Inspired by the geometric properties of diffusion models revealed through the Frenet-Serret formulas, we propose constant total rotation schedule (TORS), a scheduling strategy that ensures uniform geometric variation along the sampling trajectory. TORS outperforms previous training-free acceleration methods and produces high-quality images with 10 sampling steps on Flux.1-Dev and Stable Diffusion 3.5. Extensive experiments underscore the adaptability of our method to unseen models, hyperparameters, and downstream applications. Code is available at https://github.com/zju-pi/TORS.

cs.CV

版本更新 · v2arXiv:2606.00866

Idleness is Relative: Exploiting Tool-Call Idle Windows for Offloading in Agentic Systems with MORI

Tian Xia, Yi Xu, Hanchen Li, Zhifei Li, Xiaokun Chen, Hao Kang, Yifan Qiao, Ion Stoica

原文摘要

Agentic programs alternate between LLM inference and tool calls, each request extending the previous one's context. Program-aware schedulers pin a program's KV cache in GPU HBM across tool calls to avoid re-prefilling it at every step, but as contexts grow, pinned KV cache exceeds HBM, raising program eviction: deciding when a program releases GPU memory and when it returns. Offloading to CPU DRAM lets an evicted program wait out a tool call instead of recomputing its context, but existing schedulers leave DRAM to LRU, so offloaded programs thrash. Making these decisions is hard: tool-call durations vary by orders of magnitude and are unknown in advance. We observe that programs nevertheless stay busy (short tool calls) or idle (long ones) for extended periods, and use a program's idleness, its recent fraction of time in tool calls, as a continuous, relative metric. MORI ranks programs by idleness, keeps the least idle in HBM and the next in DRAM, and drops the rest, enforcing admission control at both tiers under any GPU-to-CPU capacity ratio; it also proactively demotes idle programs to make room for ready ones. On real coding-agent traces, MORI achieves on average 64% higher throughput and 72% lower p90 TTFT than the best offloading baseline under high load.

cs.OS

版本更新 · v2arXiv:2512.15705

Batched Early-Exit LLM Inference Done Right with DREX

Xuting Liu, Daniel Alexander, Jiaheng Lu, Siva Kesava Reddy Kakarla, Behnaz Arzani, Vincent Liu

原文摘要

Early-exit (EE) LLMs let confident tokens skip a model's deep layers, but batched serving loses most of this gain. Current systems either force tokens to exit before their own ramp is confident, an error the ramp does not bound, or save almost nothing when decoding is weight-bound, because any continuing token makes a step read the deep weights. We present DREX, a batched EE serving system that never forces a token to exit. DREX introduces cross-iteration batching: tokens that continue past the exit layer wait there, and every few steps one deep pass over all waiting tokens reads the deep weights once for several steps. Its two policies differ in what such a token's request does: holding pauses it until the deep pass, and speculative holding lets it keep decoding from the exit layer's own guess, which the next deep pass verifies. Exited tokens are trusted and never verified. Using a calibrated cost model, a controller picks speculative holding where reading weights dominates a step and guesses are often right, and holding elsewhere. DREX keeps vLLM's asynchronous scheduling by planning each deep pass from exit counts of earlier steps. With synthetic exits where 90% of tokens exit at a quarter of the depth and 40% of guesses are correct, DREX achieves 2.22-2.24$\times$ decode speedup over non-EE across batch sizes on Llama-2-70B, while per-row skipping (vSkipper's design) gains at most 7.0%. On LayerSkip models, DREX adds no statistically resolved quality loss to the EE model's own. On CNN/DM at full memory, DREX with LayerSkip achieves 1.58-1.91$\times$ decode speedup while remaining within 1.4 ROUGE-2 points of the non-EE model. These speedups are 9-63% larger than EAGLE's across the evaluated model sizes, with each method measured against its own non-EE target.

cs.DC · cs.LG

新论文 · v1arXiv:2611.14864

When Safety Becomes an Attack Surface: Manipulating Abstention in Tool-Using AI Agents

Charles Xiao

原文摘要

An AI agent's decision not to act can itself become an attack surface. We study warning-induced over-abstention, an availability attack in which adversaries place fabricated risk warnings in low-trust content to prevent legitimate task completion. Our central premise is that safe abstention requires evaluating both the claimed risk and the authority of the evidence supporting it. We develop controlled counterfactual task variants that vary risk content, source provenance, and warning wording, and evaluate a Verify-Before-Abstain protocol that corroborates warnings against authoritative records or the underlying content. Across three model families and 51 injectable AgentAbstain task pairs, caution-first prompting amplifies the attack. For Gemini, induced over-abstention rises from $9.8\%$ to $86.3\%$, despite unchanged clean-task completion. Matched wording comparisons further reveal sensitivity to warning language, with substantial variation across models and environments. Provenance-aware verification improves legitimate completion, while forgery and sensitive-content controls expose its dependence on authoritative evidence and its limitations in preserving confidentiality. These results motivate evaluating agent safety through the justification for restraint, alongside the safety of actions taken.

cs.CR · cs.AI

版本更新 · v3arXiv:2510.09133

PAC Reasoning: Controlling the Performance Loss for Efficient Reasoning

Hao Zeng, Jianguo Huang, Bingyi Jing, Hongxin Wei, Bo An

原文摘要

Large reasoning models (LRMs) have achieved remarkable progress in complex problem-solving tasks. Despite this success, LRMs typically suffer from high computational costs during deployment, highlighting a need for efficient inference. A practical direction is to switch the LRM between thinking and non-thinking modes dynamically. However, such approaches often introduce additional reasoning errors and lack statistical guarantees for the performance loss, which are critical for high-stakes applications. In this work, we propose Probably Approximately Correct (PAC) Reasoning that controls the performance loss under the user-specified tolerance. Specifically, we construct an upper confidence bound on the performance loss and determine a threshold for switching to the non-thinking model. Theoretically, using the threshold to switch between the thinking and non-thinking modes ensures bounded performance loss in a distribution-free manner. Our experiments on reasoning benchmarks show that the proposed method can save computational budgets and control the user-specified performance loss.

cs.AI · cs.LG · math.ST

版本更新 · v2arXiv:2512.18535

Integrated Control and Communication in LQG Systems

Sepehr Jahangiri, H. Ali Talebi

原文摘要

In this paper, we study the Integrated Communication and Control (ICAC) problem. Specifically, we investigate how messages can be transmitted from the controller/encoder to the observer/decoder through the control signal in Multiple-Input Multiple-Output (MIMO) vector-state Linear Quadratic Gaussian (LQG) systems under control constraints. We provide a computable capacity expression using semidefinite programming. We further show that it is possible to transmit data at a nonzero rate over an LQG system while maintaining the same optimal control cost as in the case where no information message are transmitted. Finally, we discuss how this framework generalizes communication over MIMO Gaussian channels with feedback, both with and without InterSymbol Interference (ISI).

cs.IT