跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

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

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1673新增论文
+1257其中 AI 相关
+198已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-14 02:49 · 全站最近入库：2026-11-13 20:44各信源进度 →

统计口径与数据新鲜度

统计 2026-11-13 03:21 — 2026-11-14 03:21（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：2026-11-14 02:49最近发现新增版本：2026-11-13 20:44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6,727 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

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

找到 5,505 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-14 02:48（北京时间）
版本更新 · v3arXiv:2304.03426

Convex Minimization with Integer Minima in $\widetilde O(n^4)$ Time

Haotian Jiang, Yin Tat Lee, Zhao Song, Lichen Zhang

原文摘要

Given a convex function $f$ on $\mathbb{R}^n$ whose set of minimizers is nonempty and closed, lies in a box of radius $R$ with $\log R\le\mathrm{poly}(n)$, and has integral extreme points, we give a randomized algorithm that finds an integral minimizer of $f$ with high probability using $O(n^2\log n+n\log R)$ calls to a separation oracle, $O(n+\log R/\log n)$ calls to an approximate shortest vector routine such as the LLL algorithm, and $O(n^4\log(nR)\log\log n)$ other arithmetic operations. The previous best polynomial time algorithm [Jiang, SODA 2021, JACM 2022] achieves $O(n(n\log\log n/\log n+\log R))$ oracle complexity, but its overall runtime is at least $\widetildeΩ(n^8)$. We obtain our speedup by combining a variant of the volumetric center cutting plane method of [Vaidya, FOCS 1989], run on the hybrid barrier of [Lee, Sidford, Wong, FOCS 2015] with the fast implementation of [Jiang, Lee, Song, Wong, STOC 2020], and a variant of the LLL algorithm that reduces segments of consecutive basis vectors recursively. The method runs in blocks, so that shortest vectors are computed only $O(n+\log R/\log n)$ times, and each computation takes $O(n^3\log^3 n)$ arithmetic operations in exact arithmetic. For submodular function minimization, our result implies an algorithm using $O(n^3\log n)$ calls to an evaluation oracle and $O(n^4\log^3 n)$ arithmetic operations in total. Its oracle complexity is better than that of the algorithms of [Lee, Sidford, Wong, FOCS 2015] and [Dadush, Végh, Zambelli, SODA 2018, MOR 2021], which had the best running time before this work.

cs.DS · cs.DM · math.OC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13487

The WoW: A Large-Scale World Knowledge Corpus for Full-Lifecycle LLM Training

Tianze Sun, Yongyao Duan, Sam Chen

原文摘要

Existing Chinese educational corpora often suffer from fragmented coverage and a lack of systematic organization. In this paper, we propose WoW (\textbf{Wo}rld Kno\textbf{w}ledge), a large-scale dataset designed for the full-cycle training of foundation models. The construction of this dataset is driven by an Educational Knowledge Graph, which serves as a taxonomic backbone for the systematic collection and completion of missing educational modules across all levels. Compared to previous versions, this iteration provides enhanced systematicity, superior quality, and a larger scale, while incorporating multimodal signals and supporting every stage of model development. Experiments demonstrate that WoW significantly improves model performance across diverse benchmarks. To facilitate community research, we further provide an interactive platform for graph-based data selection and retrieval.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2509.16915

Differential Privacy for Euclidean Jordan Algebra with Applications to Private Symmetric Cone Programming

Zhao Song, Jianfei Xue, Lichen Zhang

原文摘要

In this paper, we study differentially private mechanisms for functions whose outputs lie in a Euclidean Jordan algebra. Euclidean Jordan algebras capture many important mathematical structures and form the foundation of linear programming, second-order cone programming, and semidefinite programming. Our main contribution is a generic Gaussian mechanism for such functions, with sensitivity measured in $\ell_2$, $\ell_1$, and $\ell_\infty$ norms. Notably, this framework includes the important case where the function outputs are symmetric matrices, and sensitivity is measured in the Frobenius, nuclear, or spectral norm. We further derive private algorithms for solving symmetric cone programs under various settings, using a combination of the multiplicative weights update method and our generic Gaussian mechanism. As an application, we present differentially private algorithms for semidefinite programming, resolving a major open question posed by [Hsu, Roth, Roughgarden, and Ullman, ICALP 2014].

math.OC · cs.CR · cs.DS · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.18194

T-SANDHI: Tone Sandhi-aware Adaptive Network with Decoupled Hybrid Injection for Low-resource Taiwanese Hokkien Speech Recognition

Hung-Yang Sung, Chien-Chun Wang, Tien-Hong Lo, Yu-Sheng Tsao, Yung-Chang Hsu, Berlin Chen

原文摘要

In Taiwanese Hokkien automatic speech recognition (ASR), prior studies often treat tone sandhi as a major challenge under the assumption that models fail to process implicit phonological variations. However, our experiments on Taiwanese Hokkien reveal that speech foundation models actually handle tone sandhi variations effectively, and the real performance bottleneck stems from a localized confusion between these variations and retained citation tones. To address this, we propose T-SANDHI to explicitly decouple surface acoustics from underlying lexical intent on top of a frozen Whisper backbone. Using a lexicon-guided multi-task learning structure driven by text-derived pseudo labels, our lightweight hybrid injection module integrates independent citation and sandhi phonetic streams via dynamic gating. Extensive evaluation on the TAT-MOE corpus and two blind test sets demonstrates that this explicit disentanglement effectively resolves tonal mapping confusion, outperforming baselines with strict parameter efficiency.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2602.02711

Dynamic Mixed-Precision Routing for Efficient Multi-step LLM Interaction

Yuanzhe Li, Jianing Deng, Jingtong Hu, Tianlong Chen, Song Wang, Huanrui Yang

原文摘要

Large language models (LLMs) achieve strong performance in long-horizon agentic tasks through multi-step interaction and reasoning at test time. However, repeatedly invoking these models incurs substantial inference cost over an episode. Post-training quantization (PTQ) reduces this cost, but using a fixed low precision throughout an episode can sharply lower task success. We observe that interaction steps differ sharply in their sensitivity to quantization: given the same context, decoding a step at low precision turns a success into a failure at only about 1% of the steps, which we call critical steps. Based on this observation, we propose Dynamic Mixed-Precision Routing (DMR), a framework that adaptively selects between high-precision and low-precision versions of the same LLM at each decision step. Leveraging an entire-space factorization to jointly predict a step's divergence and its effect on the outcome, DMR learns to recognize these critical steps by training only a lightweight router on the model's hidden state. By reusing a shared high-precision KV cache and the hidden states, DMR avoids redundant context prefill and a separate text encoder, enabling precision switching with less than 1 ms of overhead per step and adding less than 0.11% to step latency for routing. This mixed-precision design also enables nested quantization, reducing the memory footprint by 39% on average compared with FP16. Across three long-horizon agentic benchmarks with nested quantization, DMR achieves an average 2.44x decode speedup over FP16 at comparable task success rates, without fine-tuning the underlying model. On ALFWorld with GPTQ, DMR achieves a 1.71x decode speedup with 86.9% success, compared with 87.2% for FP16.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2512.08991

Action-Guided Reachability for Deterministic World Models of Vision-Based Systems

Yuang Geng, Zhongzheng Zhang, Chengzhen Jiang, Yanru Li, Xinyang Wang, Zhuoyang Zhou, Ivan Ruchkin

原文摘要

End-to-end vision-based controllers map raw camera images directly to control actions, yet verifying their closed-loop behavior remains challenging: no analytical model maps a physical state to a camera image, and propagating a set of states through high-dimensional images yields loose reachable sets. We address both challenges with two techniques. First, we propose a Deterministic World Model (DWM), a latent-free decoder that generates camera images only from the physical state, trained with a saliency-weighted reconstruction loss and a control-consistency loss. The DWM replaces the camera in the closed loop, so a sound and tight reachable tube of the surrogate closed loop can be computed without bounding any latent variables. Second, we propose an action-guided partition-and-verify scheme that certifies as many initial sets as possible within a fixed computation budget. A Gaussian process learned from cheap simulations orders the cells and splits those that lie across the property boundary, while every other unknown cell is split along the state dimension that most changes the controller's actions along the simulated closed loop, i.e., the one that most reduces the pixel uncertainty reaching the controller. On a CARLA braking system and three Gym benchmarks, the DWM is more accurate to the real camera than a cGAN surrogate and yields 1.9x to 69x narrower robustness intervals. The vision part causes most of the remaining overapproximation, and our splitting reduces the unknown fraction of the four-dimensional CartPole initial set from 21.88% to 8.18% within the budget.

cs.CV · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.14804

AgentWeave: Efficient Distributed Agent Serving via Flow Decomposition

Kaibin Guo, Pengtu Li, Zicong Hong, Zhiyuan Fang, Wuhui Chen, Rachid Guerraoui, Anne-Marie Kermarrec

原文摘要

Agent applications are evolving into workflows composed of large language model (LLM) calls and tool invocations, termed modules. Existing distributed agent systems typically adopt a disaggregated architecture, deploying different modules on separate devices. While this enables pipeline parallelism, imbalanced module latencies introduce severe pipeline bubbles that limit throughput. A natural alternative is an aggregated architecture that co-deploys all modules across all devices, eliminating bubbles by keeping every device continuously active. However, this introduces request blocking: short requests are forced to progress in lockstep with longer ones in the same batch, and newly arriving requests must wait until the entire ongoing batch completes the full workflow. To address this, we propose AgentWeave, an efficient distributed serving system for agent applications built on the aggregated architecture. Its core mechanism, flow decomposition, decomposes long agent requests into fine-grained execution units while preserving agent semantics, allowing short and newly arriving requests to proceed without being stalled by long-running ones. Complementing this, a flow-adaptive scheduler dynamically balances the throughput gains of decomposition against its scheduling overhead. Experiments across four representative agent applications demonstrate that AgentWeave improves throughput by 1.28-1.81x, reduces average latency by 2.08-2.45x, and reduces P90 latency by 1.56-2.04x over state-of-the-art baselines.

cs.DC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13454

SafeRoute-MAS: Coverage-Aware and Risk-Calibrated Message Routing for Multi-Agent Scientific Reasoning

Zhiming Xue, Jianan Liu, Weiran Yan, Aicheng Xie, Jing Yang

原文摘要

We introduce SafeRoute-MAS, a coverage-aware, risk-calibrated procedure that keeps all specialist messages or removes one using only pre-integration information. On a sealed cache of 5,328 committees, its source-compatible operating point gains +1.80 percentage points (95% CI [+1.13, +2.52]) while removing 12.76% of message tokens. We also run 4,580 controlled integrations over a stratified 458-committee sample on a public gpt-oss-120b endpoint. The sample estimate is -0.09 points; design weighting to the original workload gives +0.94 points (95% CI [-1.31, +3.50]) and 18.72% acted-input reduction. A weighted token-matched comparison likewise does not establish a selector advantage (+1.68 points, CI [-1.27, +5.10]). Only 13 of 88 cached non-neutral labels (14.8%) retain direction by repeated-call majority. Expected-corruption UCL is 4.60% after weighting, but only 2 of 11 prespecified marginals pass the support/risk rule. We therefore propose a numerical qualification with Keep-All fallback, applied here retrospectively rather than validated prospectively.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13426

BioMem: Agentic Memory for Biomedical LLM Agents

Mingchen Li, Yongkang Xiao, Rumeng Li, Rui Zhang, Hong Yu

原文摘要

Large language models (LLMs) have demonstrated considerable potential in biomedical applications, while agentic memory systems have emerged as a promising approach to preserving and utilizing information across tasks and interactions. However, a fundamental question remains: how effectively and reliably do these systems retain, retrieve, update, and utilize biomedical information? In this study, we systematically investigate agentic memory in the biomedical domain, focusing on task effectiveness, retrieval quality, information fidelity, and long-term robustness. We evaluate five memory approaches across four LLMs in three representative settings: biomedical knowledge reasoning, clinical prediction, and long-term conversational memory. Our analyses reveal how memory construction, evidence relevance, and conversational history shape downstream biomedical performance, highlighting the distinct challenges of retrieving relevant information, preserving its fidelity, and effectively utilizing it for reasoning. Building on these findings, we introduce Hybrid Memory Retrieval and Utilization (HMRU), a training-free framework that jointly enhances evidence accessibility and task-aware utilization to improve biomedical reasoning and long-term conversational memory. HMRU achieves an average relative improvement of 8.6% across biomedical question-answering and conversational memory benchmarks, with additional gains in selected clinical prediction tasks. It further reduces conversational answer errors by 24.9% and closes 48.0% of the performance gap to full-context inference. Our findings provide a systematic characterization of biomedical agentic memory and practical insights into developing more effective and reliable memory-augmented LLM systems.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13425

Prompt2Echo (P2Echo): Text-Prompted Multi-View Echocardiography Segmentation with One Model

Moein Heidari, Alvin Kimbowa, Mahdi Tabatabaei, Afshin Bozorgpour, Sina Ghorbani Kolahi, River Jiang, Wenjin Chen, David J. Foran 等 11 位作者

原文摘要

Automated segmentation in multi-view echocardiography is essential for comprehensive cardiac assessment, yet remains challenging due to substantial anatomical and appearance variability across standard acoustic windows. In this paper, we present P2Echo, a unified framework capable of segmenting diverse cardiac structures across multiple standard echocardiographic views using a single model controlled by structured view-aware text prompts. Unlike recent foundation model adaptations that rely on implicit latent priors, P2Echo explicitly conditions the segmentation process via a novel Dynamic Image-Text Alignment (DyITA) module. DyITA integrates a differentiable routing mechanism and token differential operators to dynamically modulate decoder features according to the requested view and target structure. This allows the network to adaptively switch its processing pathways for different views without architectural changes. Extensive experiments on a large-scale internal dataset (seen during training) and held-out evaluation on three external datasets demonstrate that P2Echo matches or exceeds both view-specific specialists and state-of-the-art universal models at a fraction of their computational cost, offering an efficient, controllable solution for multi-plane cardiac analysis. Code is available at: https://github.com/moeinheidari7829/P2Echo.

cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2512.02581

Generative Online Reinforcement Learning

Chubin Zhang, Zhenglin Wan, Feng Chen, Fuchao Yang, Lang Feng, Yaxin Zhou, Xingrui Yu, Yang You 等 10 位作者

原文摘要

Reinforcement learning (RL) faces a persistent tension: policies that are stable to optimize (e.g., Gaussians) are often too simple to represent the multimodal action distributions required for complex control. Conversely, expressive generative policies---such as diffusion and flow matching---can be difficult to optimize in online RL due to intractable likelihoods and gradients propagating through long sampling chains. We address this tension with a key structural principle: decoupling optimization from generation. Building on this, we introduce GoRL (Generative Online Reinforcement Learning), an algorithm-agnostic framework that trains expressive policies from scratch by confining policy optimization to a tractable latent space while delegating action synthesis to a conditional generative decoder. Using a two-timescale alternating schedule and anchoring decoder refinement to a fixed prior, GoRL enables stable optimization while continuously expanding expressiveness. Empirically, GoRL consistently outperforms unimodal and generative baselines across diverse continuous-control tasks. Notably, GoRL achieves returns exceeding 870 on HopperStand, more than $3\times$ the strongest baseline; on high-dimensional humanoid tasks, it further outperforms the strongest non-GoRL baseline by over an order of magnitude. Code is available at https://github.com/bennidict23/GoRL.

cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2608.09886

Entanglement-assisted quantum locally recoverable codes: bounds and constructions with availability

Rutuja Kshirsagar, Gretchen L. Matthews, Julia Shapiro

原文摘要

In this work, we define entanglement-assisted quantum locally recoverable codes with availability, in which any set of up to $δ-1$ erased qudits can be recovered from any one of $t$ local recovery sets, each of size at most $r+δ-1$, with the recovery sets intersecting exactly in the erased coordinates, where $r$ is a (small) positive integer. We show that shared entanglement permits $t>1$, meaning that multiple local recovery sets can be available for the same set of up to $δ-1$ erasures. We establish a Singleton-like bound for this family of codes and present random constructions based on classical linear codes with Vandermonde parity-check matrices. We also provide explicit constructions of entanglement-assisted quantum locally recoverable codes with availability from several classical code families and their folded versions, including Tamo-Barg codes, fiber-product codes, and algebraic-geometry codes such as one-point Hermitian and Suzuki codes.

cs.IT · quant-ph

查看论文
arXiv ↗PDF ↗
版本更新 · v5arXiv:2601.21343

Self-Improving Training & Thinking

Ellen Xiaoqing Tan, Jack Lanchantin, Zichun Yu, Polina Kirichenko, Shehzaad Dhuliawala, Danwei Li, Thao Nguyen, Jing Xu 等 14 位作者

原文摘要

Large language models are classically trained in stages: pretraining on raw text followed by post-training for instruction following and reasoning. However, this separation creates a fundamental limitation: many desirable behaviors such as safety, factuality, overall generation quality, and reasoning ability are only added at a late stage, even though the patterns learned earlier strongly shape a model's capabilities. To tackle this issue, we introduce a new way to pretrain and mid-train models that incorporates these behaviors earlier. We utilize an existing strong, post-trained model to both rewrite data and to judge policy model rollouts, thus introducing thinking and reinforcement earlier in training. We further improve reasoning at all stages by introducing the concept of self-improving post-thinking: improving the model's reasoning and self-judging ability by reasoning after generation. In our experiments, we show these approaches can give strong gains across multiple axes.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13393

Plumbline: When the Loss Lies, Validating LLM Training Correctly

Ruochen Wen, Hongyang Wang, Siwei Li, Wenbo Chen, Xian Li, Jingguo Liu, Leo Fan, Yiming Yan 等 11 位作者

原文摘要

Correctness is a prerequisite for training large language models (LLMs) at scale, yet it is hard to ensure: a fault can sit in any part of a run, and how it shows differs across frameworks and the libraries they depend on. Some failures are loud and some are silent, and both take time to detect and trace. A single fault can waste a long run: the run crashes late, or the model scores lower only at the final evaluation. We present Plumbline, a correctness validation system for LLM training, used in production at Amazon. Plumbline checks each part of the training lifecycle, from the data a run trains on to the model it serves, down to the new layers each model brings and up to whole runs. Its diagnosis reuses these checks to trace failures to their root causes. Both rest on a systematic correctness methodology designed to carry across models and frameworks. Plumbline supports two agentic use cases: Day-0 onboarding, which validates the training of a newly released model before teams rely on it, and root-cause tracing of training failures. Across 3 frameworks and 5 models, it found 61 problems, 32 of them silent. Of these, 42 are fixed or reported upstream, 8 of them first raised by us. We further group the 61 into seven failure patterns that show where training breaks, and report how problems recur across frameworks and models.

cs.DC · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13387

Towards Accessible, Autonomous Photosynthesis Monitoring on Living Leaves

Qitong Li, Xiaoyu Zheng, Xuanang Chen, Julia Wang, Raj Nileshbhai Dave, Dayanara Chavez, Nirvik Prakash Iyer, Nivedita Arora

原文摘要

Photosynthesis is a direct biomarker of plant health and functioning. However, measuring net photosynthesis on individual leaves requires expensive (USD 42,000 - USD 89,000) instruments and manual effort. This limits access to photosynthesis sensing and makes long-term monitoring across multiple leaves difficult. We present a 1000x low-cost, autonomous sensing system that enables longitudinal monitoring of photosynthesis on living leaves over multiple days. Guided by a formative user study with 13 plant biology researchers, the system integrates an automated, leaf-attached chamber with a low-cost CO2 sensor. Built using a 3D-printed clamp and thermoformed dome, the chamber periodically closes to measure CO2 drawdown and reopens to refresh the air. Each chamber costs about USD 62 and operates independently, allowing multiple leaves to be monitored simultaneously without pumps or gas tubing. We characterize each chamber and validate its measurements against a research-grade gas exchange instrument (CIRAS-4) across 29 sessions (R^2 = 0.90). We also demonstrate monitoring of C3 and CAM photosynthesis plants, adaptability to different leaf geometries, and a light-response experiment. In a nine-day deployment, six chambers captured simultaneous CO2 exchange from individual leaves of the same plant during drought and rewatering. Each leaf showed different patterns of nighttime CO2 uptake and recovery, which a whole-plant CO2 measurement instrument would obscure. To our knowledge, this is the first demonstration of such leaf-level photosynthesis dynamics using independent, low-cost autonomous chambers. Our work opens new opportunities for plant science, education, and data-driven agriculture.

cs.HC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13370

AHEADance: Anticipatory Humanoid Dance from Streaming Audio

Tianhu Peng, Yukun Wang, Lingfan Bao, Yiyang Jia, Chengxu Zhou

原文摘要

A humanoid that dances to streamed music must commit each motion chunk before it has heard the music that chunk accompanies, and must continue from motion it has already committed. We present AHEADance, a robot-native streaming dance generator executed on a humanoid robot by a whole-body tracker. ListenAhead conditions each chunk on heard music and a frozen music model's forecasts of the coming music, aligned to their target times and weighted by reliability. The forecasts carry the timing of upcoming beats almost as well as the actual future; with them, the generator is better synchronized with the music and concentrates more motion energy at its song's beat frequency. Commit Forcing trains continuation from the model's own committed chunks, updating both the latent history and the decoded boundary of the next chunk; it reduces low activity from 83.5% to 10.3%, the level of real dance. Three music evaluators, each validated on real G1 dance, test synchronization, song and style. On a physical Unitree G1, it dances while the music plays and comes to rest when it stops, without a fall. Project page: https://ihcr.github.io/AHEADance

cs.RO · eess.SY

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13368

Ender: Visual Scaffolding for Geometric Proof Comprehension

Hwei-Shin Harriman, Yuchen Jin, Dominik Moritz, Joshua Sunshine

原文摘要

Geometric proof is a foundational but notoriously difficult topic in secondary mathematics, primarily due to the high cognitive load required to integrate visual diagrams with logical text. As proofs evolve, tracking the initial givens and an expanding set of derived truths can easily overwhelm working memory. While established scaffolding techniques are often visual and diagram-based, current digital learning environments mostly rely on static, text-heavy interfaces that force students to manage these complex mental models on their own. This paper introduces Ender, an interactive rendering system designed to visualize the logical status of two-column proofs. Ender's design is grounded in four literature-derived requirements: (1) diagram annotation, (2) structural tracing, (3) support for partial or incorrect work, and (4) synchronization between the diagram and text. To evaluate the system, we conducted a between-subjects, think-aloud experiment with 12 secondary school students in the United States, comparing Ender against a traditional digital proof interface. Our results indicate that students using Ender performed significantly better on proof comprehension tasks--specifically in identifying and correcting logical errors. These findings suggest that dynamic visual scaffolding may ease the working memory burden of digital interfaces, offering a promising building block for future AI-based proof tutoring systems.

cs.HC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13348

ProTEUS: A mutation-resolved world model for closed-loop protein evolution

Yikun Zhang, Minghao Guo, Guangxue Xu, Peter Yichen Chen, Wojciech Matusik, Wengong Jin, Jingjing Chai, Kaiyi Jiang

原文摘要

Protein engineering requires navigating rugged fitness landscapes, where small sequence changes can produce large functional differences, but experimental screening can only sample a small fraction of sequence space. Thus, a central challenge is to efficiently traverse the large combinatorial space of mutations in silico. We introduce ProTEUS, a mutation-resolved world model for closed-loop protein evolution. Pretrained on deep mutational scanning assays, ProTEUS learns a joint representation of sequence, evolutionary, and structural information using MutationFormer, a mutation-biased attention architecture. Variants are represented as states in this joint latent space, and mutations define actions whose functional consequences are predicted. To support exploration of many mutants, we developed a fast, physics-constrained all-atom predictor that utilizes explicit geometric constraint enforcement to achieve >10 fold increase in speed than AlphaFold3 with guaranteed physical validity. Lastly, ProTEUS refines its prediction and understanding of protein activity through closed loop experimental feedback to guide evolution toward a desired phenotype. ProTEUS surpasses established models in both zero-shot and supervised fine-tuning mutation prediction tasks. We experimentally demonstrate its utility in two closed-loop multi-objective engineering campaigns on aldehyde dehydrogenase and levansucrase, yielding variants with >10 fold higher activity than wild type. ProTEUS provides a world-model framework that connects virtual evaluation of mutational outcomes with experimental learning for efficient protein evolution.

q-bio.BM · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2601.05437

Tracing Moral Foundations in Large Language Models

Chenxiao Yu, Bowen Yi, Farzan Karimi-Malekabadi, Suhaib Abdurahman, Jinyi Ye, Shrikanth Narayanan, Yue Zhao, Morteza Dehghani

原文摘要

Large language models often produce human-like moral judgments, but it is unclear whether this reflects an internal conceptual structure or superficial ``moral mimicry.'' Using Moral Foundations Theory (MFT) as an analytic framework, we study how moral foundations are encoded, organized, and expressed across 14 base and instruction-tuned LLMs spanning four model families (Llama, Qwen2.5, Qwen3-MoE, Mistral) and scales from 7B to 70B. We employ a multi-level approach combining (i) layer-wise analysis of MFT concept representations and their alignment with human moral perceptions, (ii) pretrained sparse autoencoders (SAEs) over the residual stream to identify sparse features that support moral concepts, and (iii) causal steering interventions using dense MFT vectors and sparse SAE features. We find that models represent and distinguish moral foundations in a manner that aligns with human judgments, and that this moral geometry naturally emerges from pretraining and is selectively rewired by post-training. At a finer scale, SAE features show clear semantic links to specific foundations, suggesting partially disentangled mechanisms within shared representations. Finally, steering along either dense vectors or sparse features produces predictable shifts in foundation-relevant behavior, demonstrating a causal connection between internal representations and moral outputs. Together, our results provide mechanistic evidence that moral concepts in LLMs are distributed, layered, and partly disentangled, suggesting that pluralistic moral structure can emerge as a latent pattern from the statistical regularities of language alone.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2611.06858

CompressionBench: How can your coding agents compress LLMs?

Xinrui Zhong, Zhifeng Li, Zhiming Wang, Xinze Feng, Yongqi Liu, Shanglin Yang, Yang Zhou, Zikun Li 等 10 位作者

原文摘要

Recursive self-improvement requires AI systems to improve the models and infrastructure used by subsequent systems. LLM compression offers a concrete efficiency objective: retain a pretrained model's behavior within a smaller model storage budget. We study whether coding agents can carry out this step autonomously. We formulate compression as a terminal task in which an agent receives a source checkpoint, storage limit, inference environment, and time budget, then constructs compressed weights and executable inference code. The agent controls its algorithm and experiments; perplexity feedback guides search, and independent verification checks the final model. Across six agents and six model--size settings, 29 of 36 submissions pass verification. Selecting the lowest-perplexity agent design per task yields an average 3.4% perplexity increase over BF16 and 77.33% downstream accuracy, compared with 79.56% for BF16. Recorded experiments reveal how agents achieve these outcomes and where improvement breaks down. Agents lower PPL in 21 of 24 revisions at nearly fixed storage; complete-model feedback reverses 17 of 79 prior local rankings; and generation and inference tests expose failures that likelihood alone misses. The results identify representation design, experimental judgment, and capability validation as central to agent-driven model compression, a measurable component of efficiency-oriented self-improvement.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13336

Veritas: A Verifier Generation Harness for Agent Tasks

Daoan Zhang, Qinbo Bai, Dehai Min, Xingyu Mu, Ziyi Chen, Jianing Yin

原文摘要

The effectiveness of a self-improving system relies heavily on the quality of its verifier, which judges the outcome of each improvement round. However, the verifiers of most tasks are written by hand, which is expensive and does not scale to tasks that come without one. We propose Veritas, a harness that generates verifiers automatically. Unlike general-purpose agent harnesses, Veritas adds three mechanisms for verifier generation: an evidence header stating what the verifier will observe, a library of domain rules that tasks leave unstated, and authoring principles, partly refined by a self-improvement loop. We evaluate on six benchmarks: four scored by execution, whose official verifiers we first certify, and two that compare pairs of responses. With the model fixed, the verifiers Veritas generates agree better with the official verdicts than those of seven other harnesses on all four execution-scored benchmarks. On the two pairwise benchmarks they are at or near the best. They also agree better than those of the vendor's own harness for two of three other model families. Its mechanisms also improve 9 of 12 harness-benchmark pairs scored by execution across three harnesses.

cs.AI · cs.CR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13334

Disentangling Multimodal Rotary Positional Embedding in Vision-Language Models

Zhaoyi Liu, Haichuan Zhang, Xiaochuan Gong, Yancheng Zhu, Bowei Tian, Wanghao Ye, Yufan Zhang, Ang Li

原文摘要

Vision-language models widely adopt Multimodal Rotary Position Embedding (MRoPE), an extension of Rotary Position Embedding (RoPE) that incorporates temporal, height, and width axes to encode the spatial structure of visual tokens. However, MRoPE also assigns height and width coordinates to text tokens that lack inherent spatial structure, introducing spatial positional effects into text-vision attention. Existing MRoPE variants explore different frequency allocations and coordinate layouts, with their designs suggesting different intuitions about these cross-modal effects: they may provide useful spatial references for associating language with visual regions, or introduce unnecessary positional preferences that hinder text-vision interactions. In this work, we disentangle intra-visual spatial structure from cross-modal spatial effects to investigate this question. We identify a constraint of shared coordinate layouts that prevents eliminating cross-modal spatial effects while preserving intra-visual spatial encoding, and develop controlled experiments to isolate their respective contributions. Our results across multiple model scales, MRoPE variants, and benchmarks show that cross-modal spatial effects provide little additional benefit as spatial references and can even impair model performance, while intra-visual spatial encoding remains useful for fine-grained grounding.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2509.22979

OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning

Xinzhi Zhang, Zeyi Chen, Humishka Zope, Hugo Barbalho, Konstantina Mellou, Marco Molinaro, Janardhan Kulkarni, Ishai Menache 等 9 位作者

原文摘要

Mathematical optimization drives decisions in supply chains, logistics, energy, and scheduling, but translating natural-language problems into solver-executable formulations remains a bottleneck. While frontier models show promise for this task, many deployments require small language models (SLMs) that run locally under privacy, latency, and cost constraints. Existing SLMs are limited by noisy data, ambiguous benchmarks, and one-shot inference that ignores the iterative, feedback-driven nature of optimization modeling. To address these challenges, we introduce OptiMind, a system that combines data cleaning and class-specific hints with training using solver feedback. OptiMind distills recurring formulation errors into class-specific hints that guide the repair of ambiguous problems and the regeneration of solution trajectories. This process provides reliable training data, which is then used for multi-turn RL to teach models to revise their formulations in response to solver feedback. At inference time, the same hint library guides further refinement as models revise their solutions over successive turns. To support reliable evaluation, we clean three challenging benchmarks from the literature. On these cleaned benchmarks, our 20B OptiMind-RL model outperforms strong open-source baselines and reaches frontier-model performance under the same protocol. The full hint library, cleaned test data, cleaning logs, and evaluation code are publicly available on Github (https://github.com/microsoft/OptiGuide/tree/main/optimind), and our trained OptiMind-RL model is available on Hugging Face (https://huggingface.co/microsoft/OptiMind-RL).

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13292

Large Language Model Explainability for Planning Lunar Rover Trajectories under Uncertainty

Michelle Ho, Maximilian Adang, Zachary Booth, David E. Astle, Margaret Hansen, Edward Balaban, Mykel J. Kochenderfer

原文摘要

Planning planetary rover trajectories under probabilistic uncertainty is a multi-objective optimization problem that science teams currently iteratively solve and verify by hand. This process becomes impractical when timelines and constraints frequently shift throughout the mission planning phase. Sampling-based planners such as Monte Carlo tree search can speed up this process, but their outputs do not preserve traceability to the science objectives that justify each plan, which the science team needs to verify. We propose an interactive assistant based on a large language model (LLM) that can explain planner trajectories using domain vocabulary and re-run the planner under user-specified conditions to compare alternatives. Previous work has used LLMs inside planner decision-making loops, but we explore using them for post-planning explanation, where they are more likely to be trusted since they do not directly influence planner decisions. Our framework pairs a simulator inspired by the NASA Volatiles Investigating Polar Exploration Rover (VIPER) mission with a ReAct queryable interface that calls curated action tools and retrieval functions. We evaluate the assistant on a query bank consisting of plan comparisons, feasibility, and lookup queries. Experiments show that 90.9% of the assistant's responses are faithful to the planner's outputs, a nearly 65% improvement over zero-shot prompting of the same LLM without tool access, demonstrating the potential for trustworthy explainability of uncertainty-aware traverses with an LLM-based system.

cs.RO

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13252

Splitting for Stability: a Modified Reward Scheme for Transaction-Fee-Only Regimes

Eric Brigham, Baigang Chen, Nicholas Hopper

原文摘要

As Bitcoin's block subsidy declines over time, miner revenue will come to depend increasingly on transaction fees. In fee-dominated regimes, winner-take-all reward structures can create incentives for undercutting and fork-based mining strategies, leading to instability and reduced transaction throughput. To prevent this instability, we propose a reward-splitting mechanism in which the winning block producer receives an $α$-portion of the block reward, while the rest is distributed among other miners; this incentivizes mining on previous blocks with higher rewards, while preserving the incentive to produce blocks with high rewards. We evaluate the proposed reward structure through analysis and simulation. Our analysis derives a local profitability condition for one-step Function-Fork deviations, showing how the winner-side reward, loser-side subsidy, and honest hash power jointly determine whether undercutting is profitable. Simulations show parameter regions in which honest behavior dominates fork-based deviations. In the cases where forking becomes dominant, the learned strategy is the least aggressive variant studied. The analysis and simulations suggest that suitable reward-splitting parameters can reduce the honest participation needed to suppress aggressive forking strategies relative to the winner-take-all structure studied in prior work. Finally, we show that the modified reward scheme does not appear to worsen selfish mining incentives. Together, these results suggest that reward redistribution can mitigate fee-driven fork incentives without introducing stronger selfish mining incentives, providing a possible direction for incentive design in fee-era proof-of-work systems.

cs.CR · cs.GT

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13238

STAR: Semantic-Criticality Tiered Allocation of Radio Resources via Zero-Shot Learning

Albert Shaju, Christo Kurisummoottil Thomas, Mingzhe Chen

原文摘要

In remote inference, edge devices transmit high-resolution images over wireless links to a receiver, where an artificial intelligence (AI) model answers task-specific questions about these images to enable reliable decisions. These answers are often determined by a small set of task-causal pixels, such as those of text, whose corruption changes the answer, so these pixels must be delivered with high fidelity. Conventional transmission protects every pixel equally, so it either consumes many resource elements (REs) or corrupts these regions at low signal-to-noise ratio (SNR). To overcome these limitations, in this paper, a zero-shot, domain-agnostic framework, termed semantic-criticality tiered allocation of resources (STAR), is proposed. In particular, a per-pixel semantic-criticality indicator (SCI) is derived from the attention of a frozen vision-language model, so new tasks and domains are handled by changing only the prompt, without retraining. The resulting task causality-aware resource allocation problem over protection tiers is solved exactly by a threshold rule on the SCI, realized by a coded tier for the task-causal pixels and an uncoded, downsampled base layer for the full image. Simulations over a 5G New Radio link demonstrate that STAR attains a semantic accuracy of at least 0.98 for $E_s/N_0 \geq -4$ dB and, at comparable semantic reliability, reduces the REs of robust coded transmission by 59.5% and 91.5% on factory- and vehicle-monitoring datasets, respectively, the latter without any change to the models. Although a deep joint source-channel coding baseline uses fewer REs at comparable semantic accuracy, it must be retrained with labeled data for every new task or domain.

eess.SP · cs.AI · eess.IV

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.17326

Evaluating Long-Horizon Robot Planning under Uncertainty

Harsh Muriki, Gayathri Aravindan, Yuchen Zheng, Sonia Chernova

原文摘要

Household robots must successfully navigate im-perfect circumstances, both in their internal models and the external world. We introduce a benchmark of seven long-horizon household tasks in Habitat that vary initial robot memory and environment configurations. We evaluate adaptations of ReAct, VLM-TAMP, and Tru-POMDP within a shared execution framework across 90 tasks-environment-memory combinations, totaling 1,350 episodes. Our evaluation measures task success, partial completion, and action efficiency, along with rule-based analysis of execution traces. ReAct achieves the highest overall task success (45.3%) and completion (71.7%), while having accurate initial memory does not consistently yield the highest performance for every planner. We identify recurring placement errors, precondition errors, and redundant actions through trace analysis. These findings highlight differences in how the evaluated planners use initial knowledge and respond to uncertainty.

cs.RO

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13218

Wrong the Same Way: Cross-Prompt Sharing Both Suppresses and Amplifies Reward Hacking in RLVR

Michael Li, Rohan Pandey, Ray Chen

原文摘要

Two accounts of how reinforcement learning with verifiable rewards (RLVR) interacts with an imperfect verifier coexist: in one, reward hacking is inherited from the base model; in the other, RL amplifies it. We study both in a controlled environment small enough to compute the entire response distribution exactly. We vary the universality of a leak, or the fraction of prompts that accept the same wrong response, and the verifier's static false-positive rate. At its KL-regularized optimum, RLVR acts as a filter that reweights the accepted set without altering its composition, so the hack rate is inherited exactly. Taking that as a null model, a known decomposition writes the drift from it as a sum over prompts of covariances between reward and the empirical tangent kernel. The terms from other prompts vanish without parameter sharing. With sharing, every prompt that accepts a wrong answer pushes it up and every prompt that rejects it pushes it down through the baseline, in proportion to the model's probability of producing it there. A shared-parameter policy thus drives a leak that few prompts accept below the base rate and amplifies one that most prompts accept. At every convergent learning rate under Adam, SGD, and sampled GRPO, hacking ends higher at high universality than at low, but whether either end crosses the base rate depends on the step size. Independent per-prompt copies remove the dependence. The direction of this push depends on neither the verifier's error rate nor universality alone, but on how much of the model's kernel-weighted probability of a wrong answer falls on prompts that accept it, which frozen base model controls support. Verifier fixes should target the mistakes the model repeats across prompts: two patches that remove equally many errors differ by up to 38 points of true accuracy. We propose this alignment as a hypothesis for why earlier studies disagree.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13209

ARoute: Task-Aligned Batch and Streaming Model Routing

Daofu Zhang, Qingcheng Zeng, Chunwei Liu, Fuheng Zhao

原文摘要

Language models increasingly power data-intensive applications, from semantic data processing to question answering and agentic workflows. Using frontier models for every request is prohibitively expensive, while relying only on smaller models reduces accuracy. We study model routing under a user-specified cost budget, aiming to maximize accuracy under task-specific correctness criteria. Existing routers fall short in two ways: pretrained routers rely on generic preference signals that may not reflect a task's own correctness criteria or a new model pair, and model cascades run the weaker model first on every request, adding sequential latency. We present ARoute, a task-aligned router that learns cost and accuracy signals to select a model for each request before inference. ARoute has two variants: ARoute-Batch operates on a batch of requests known in advance, routing requests with the greatest expected benefit to the stronger model subject to the budget, whereas ARoute-Stream processes requests online, using a threshold calibrated on a small validation set to satisfy the cost quota and dynamically adjusting it as the budget is consumed. For ARoute-Stream, we provide a finite-sample guarantee on expected normalized spending. We evaluate ARoute on MMLU-Pro, LiveCodeBench, SuperGPQA, and two value-filling tasks built from SWAN semantic query benchmark against other state-of-the-art approaches. At matched cost, ARoute-Batch achieves the highest mean accuracy in all 15 task-cost settings, improving over the strongest baseline by 4.1 percentage points on average and by up to 11.9 points. ARoute also completes a batch up to 2.8x faster than cascades.

cs.DB

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.13188

One-Sided Expert-Parallel All-to-All for Mixture-of-Experts Inference over NVLink

Bo Li, Chulian Zhang, Or Biri, Anthony Chang, Anton Korzh, Noam Duvdevani, XingFei Xi, Daniel Stokes 等 13 位作者

原文摘要

Expert parallelism (EP) shards the experts of a Mixture-of-Experts (MoE) layer across GPUs, so every MoE layer requires two all-to-all communications, namely dispatch and combine. Compared with training, inference processes relatively few tokens per GPU at a time, and at small batch sizes fixed costs that do not grow with the number of tokens, rather than data movement, dominate the latency. We present NVLinkOneSided, a one-sided EP all-to-all for GPUs within a single NVLink domain. Each source rank pushes its tokens, without first exchanging metadata, into rank-major buffers on the target GPUs, and the grouped GEMMs read them in place instead of from a per-expert copy. We provide two protocols for the synchronization that the communication requires, namely Fence and Sentinel, for high-throughput and low-latency scenarios respectively. On Blackwell GPUs, NVLinkOneSided completes dispatch and combine 1.2-3.0$\times$ faster than DeepEP, with the largest speedups at small batch sizes. The resulting MoE layer is up to 2.4$\times$ faster than DeepEP with DeepGEMM. It even outperforms the fused DeepGEMM MegaMoE at up to 128 tokens per GPU, making it still competitive for decoding. We expose NVLinkOneSided through a modular interface, so that it is agnostic to the activation quantization format, decoupled from the MoE compute backend and easy to integrate into inference frameworks.

cs.DC · cs.CL · cs.LG · cs.PF

查看论文
arXiv ↗PDF ↗