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

+406新增论文
+306其中 AI 相关
+55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:2026-11-16 11:42各信源进度 →

统计口径与数据新鲜度

统计 2026-11-15 11:43 — 2026-11-16 11:43(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:2026-11-16 11:42库内论文最新更新时间:2026-11-14 01:29

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。

23,754 已收录论文19,886 AI 相关6,892 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 38 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)
新论文 · v1arXiv:2611.15165

Bridging Update Coordination and Flatness in Federated Learning

Chongwei He, Shichen Huang, Jiayun Li, Deyuan Lin, Wangze Jiang, Shengjia Li, Xu Wang, Xiaofan Li 等 9 位作者

原文摘要

Federated learning combines complementary knowledge across clients to alleviate local data limitations without sharing raw samples. However, data heterogeneity can cause conflicting updates and hinder generalization. Update coordination reduces interference among client updates, while flatness optimization reduces sensitivity to parameter perturbations to improve generalization. Existing joint designs largely target specific algorithms, limiting flexible combinations of these approaches. To address this problem, we propose a unified federated learning framework that bridges update coordination and flatness optimization. Clients separately accumulate descent signals and flatness corrections along the same local trajectory. The server coordinates both groups with independent states before fusing their outputs into a global update. Explicit conversion between trajectory signals and model updates enables different coordination strategies to work with compatible local or globally guided flatness mechanisms without changing the inference architecture. Experiments show that the evaluated combinations generally improve average and worst-client accuracy over update coordination alone, demonstrating the effectiveness of the proposed framework in combining these complementary optimization mechanisms.

cs.CV

版本更新 · v3arXiv:2511.17044

Parametric Retrieval-Augmented Generation using Latent Routing of LoRA Adapters

Zhan Su, Fengran Mo, Jinghan Zhang, Yuchen Hui, Jiaao Sun, Jian-yun Nie

原文摘要

Parametric Retrieval-Augmented Generation (PRAG) is a RAG approach that integrates external knowledge directly into model parameters using a LoRA adapter, aiming at reducing the inference cost compared to traditional RAG. However, current PRAG approaches adopt a \textit{one-to-one} document encoding scheme, using a dedicated LoRA adapter for each individual document. This scheme introduces two major limitations: 1) As the number of documents increases, there will be a prohibitive cost for training and storage. 2) The LoRA adapters may largely overlap due to the shared knowledge across documents, making the approach highly inefficient. To overcome these challenges, we propose the Poly-PRAG approach, which uses a small set of LoRA adapters that are able to encode more general knowledge. Each document can be encoded using a combination of them through a latent routing function. By jointly training the LoRA adapters and the latent routing function, each LoRA adapter is able to encode a shared part of the knowledge across documents, and the routing function can select the best combination of adapters for a document. Experimental results on four benchmarks demonstrate the effectiveness of the Poly-PRAG compared to other strong PRAG baselines. In addition, this approach reduces the storage requirement by avoiding the need to store a large number of LoRA adapters and offers a more efficient way to encode external knowledge into LLMs.

cs.IR

新论文 · v1arXiv:2611.12434

Frontier AI Safety and Governance 2026: Risk Understanding and Coordinated Governance

Zeng Yi, Seán Ó hÉigeartaigh, Cao Gongce, Lu Enmeng, Chen Yu, Han Kaiyu, Li Kai, Wang Zhengqi 等 110 位作者

原文摘要

Frontier AI Safety and Governance 2026: Risk Understanding and Coordinated Governance is a key outcome of the AI Safety and Governance Program of World Internet Conference Specialized Committee on AI in 2026. Drawing on the latest developments in frontier AI capabilities and applications, the report systematically examines the evolving risk landscape, analyzing key risk areas, their severity and response priorities. Building on a review of existing global governance practices and an assessment of frontier AI governance readiness, it identifies key governance gaps and weaknesses and proposes directions and practical mechanisms for coordinated responses. The report provides a reference for government departments, international organizations, businesses, research institutions and other stakeholders. It aims to strengthen the international community's understanding of and consensus on frontier AI risks, enhance preparedness and collective response capacity, and advance the safe, reliable and responsible development of frontier AI to better serve the common well-being of humanity.

cs.CY

版本更新 · v3arXiv:2606.29858

Smooth Scaling Laws Hide Stepwise Token Learning

Pingjie Wang, Zechen Hu, Peiru Yang, Fu Guo, Debing Zhang

原文摘要

Language model loss follows remarkably regular scaling laws over model and data size, yet it remains unclear why the aggregate loss should exhibit a power-law form. Existing explanations often attribute this regularity to a heavy-tailed spectrum of pattern difficulty in natural language, but this view has not been directly validated at token-level granularity in large-scale real-data training. We present a token-level framework that decomposes scaling laws into localized learning events of individual contextualized tokens. By fitting token loss trajectories with sigmoids, we show that token learning is concentrated in localized transitions, giving rise to a learning-time spectrum that dominates the scaling-law shape. Across more than one hundred pre-training runs on large and diverse real-language corpora with modern LLM architectures, scaling up to 6B parameters and 300B training tokens, the measured learning-time spectrum quantitatively reconstructs the validation loss derivative along the training-step $T$, data-scale $D$, and model-scale $M$ axes. We further show that the same signal is actionable: by reshaping the training distribution according to when tokens become learnable, we alter the optimization trajectory and achieve 11\% faster validation-loss reduction. These results provide direct empirical evidence that scaling laws are governed primarily by the distribution of token-level learning times, and that this distribution can be used not only to explain scaling behavior but also to improve training performance.

cs.CL

版本更新 · v2arXiv:2610.34045

Minimax Regret for Sobolev Kernel-Based Reinforcement Learning

Haichen Hu, David Simchi-Levi

原文摘要

We study the minimax regret rate for episodic kernel-based reinforcement learning, where the expected reward and transition density belong to a Sobolev reproducing kernel Hilbert space (RKHS), with continuous state and action spaces, and adversarial initial states. For Sobolev smoothness parameter $ν>0$, state dimension $d_S$, and action dimension $d_A$, we prove that the minimax expected regret over $T$ episodes is $\widetildeΘ(T^{\max\{e_A,e_B\}})$, where $e_A=\frac{ν+d_A}{2ν+d_A}$, $e_B=\frac{ν+d_A+3d_S/2}{2ν+d_A+2d_S}$ with the other model parameters fixed. The two exponents reflect the distinct difficulties of action selection and learning across states. We introduce Sobolev Optimal Value Iteration (\texttt{SOVI}), which attains this rate from observed rewards and transitions. For fixed smoothness and reference-kernel lengthscale, its regret bound has a prefactor polynomial in the state-action dimension $d=d_S+d_A$, horizon $H$, reward and transition RKHS norm bounds $B_r,B_p$, and the norm-comparison factor $Λ$, up to logarithmic factors; $Λ=1$ for normalized Matérn kernels. The algorithm separates initial action selection from continuation-value estimation and combines local polynomial approximation with the shared RKHS structure of the approximation errors. Our matching upper and lower bounds determine the optimal dependence on the number of episodes for Sobolev kernel-based reinforcement learning regret minimization. Thus, we partially address the open problem in Vakili(2024) under Sobolev kernel RL.

cs.LG · math.ST · stat.ML

新论文 · v1arXiv:2611.08649

VoluCompose: Constructing Compact Volumetric Visual Representations for Medical Image Understanding

Zanting Ye, Yunrui Lin, Shengyuan Liu, Jing Hao, Chen Hu, Yefeng Zheng, Lijun Lu

原文摘要

Medical vision--language models can interpret volumetric images, but processing full volumes is costly because hundreds of slices produce long, redundant visual-token sequences. Existing compression methods often couple content selection with budget determination. We observe that neighboring slices share visual features and that attention concentration varies across visual encoders. We propose VoluCompose, a training-free, model-agnostic framework that separates what visual content to retain from how many tokens to allocate. Its hyperbolic token composition module groups contiguous slices into ordered support units and uses the Lorentz distance to account for feature magnitude and direction during token allocation, anchor selection, and refinement. Its capacity control module uses visual-encoder attention entropy to allocate more tokens when attention is more dispersed. We evaluate VoluCompose on two tasks across three datasets and four backbones. Across nine calibrated settings, VoluCompose attains the highest mean quality--retention hypervolume among the compared methods. It achieves a mean relative quality of 121.68\% against the uncompressed references while retaining 22.78% of visual tokens on average. Compression remains robust even at a 10% token budget. Across settings, the lowest ROUGE-L ratio to the uncompressed reference is 96.4\% for VoluCompose. The corresponding minima for the compression baselines range from 15.0% to 70.5%. Code is available at https://github.com/yezanting/VoluCompose.

cs.CV · cs.AI

版本更新 · v2arXiv:2608.27969

openJiuwen: Beyond Static Harnesses for Long-Horizon Coding Agents

Tao Yu, Xinyu Zhang, Qianqian Chen, Xiaoneng Xiang, Chia Kwangyang, Xingchen Huang, Ran Chen, Yangkai Ding 等 19 位作者

原文摘要

Long-horizon coding agents operate over evolving repository states while increasingly relying on heterogeneous capabilities, delegated agents, and multi-agent coordination. These trends pose two complementary challenges for the agent harness. First, developers need to compose capabilities, reconfigure execution logic, and scale increasingly complex agent systems without repeatedly rebuilding orchestration. Second, complex coding tasks continuously produce new evidence---such as semantic diagnostics, execution outcomes, task progress, and changing context relevance---that should dynamically influence subsequent runtime decisions. We characterize these challenges as Structural Composability and Runtime Adaptivity. We present openJiuwen, an open-source harness designed for both developer composability and adaptive task execution. openJiuwen provides a shared execution substrate and Rail-based capability composition across single agents, delegated sub-agents, and Swarm Flow, enabling developers to construct sophisticated agent harnesses under common execution semantics. It further adapts framework-controlled runtime decisions around a fixed model policy, allowing evolving evidence to dynamically affect context, feedback, and task control toward successful completion. We systematically evaluate openJiuwen on SWE-bench Verified and Terminal-Bench 2.1, where it achieves 82.6% and 87.19%, respectively, exceeding the strongest selected official-leaderboard point estimates by 3.4 and 3.39 percentage points. These results show that openJiuwen achieves strong performance on complex coding tasks while providing a composable and adaptive harness design.

cs.AI

版本更新 · v2arXiv:2606.31833

BRIDGE: Lazy-Move Compilation for Dual-Species Neutral-Atom Quantum Computers

Chen Huang, Jingbo Wang, Zhemin Zhang, Ming Zhong, Zhuo Fu, Zhiding Liang, Yuan Sun, Dong E. Liu

原文摘要

Neutral-atom quantum computers commonly connect distant qubits by transporting atoms, incurring trap-handover errors, heating, and atom loss. We present BRIDGE (Buffer-Relay Interconnect for Data-stable Gate Execution), a hardware--software co-design that connects stationary data atoms through a fixed, second-species buffer network. Short buffer chains mediate remote controlled-Z gates and return to their initial states after each operation. BRIDGE-F combines relay routing with buffer-mediated SWAPs to relocate quantum states without transporting atoms. BRIDGE-H additionally permits data-atom transport when its full modeled cost is offset by higher survival-weighted fidelity. Across 21 circuits, BRIDGE-F exceeds ZAP and Enola in modeled survival-weighted fidelity, with geometric-mean ratios of $2.1\times$ and $38\times$ and medians of $1.3\times$ and $3.7\times$. Modeled execution latency is lower by factors of approximately $500$ and $1{,}350$, respectively, with circuit single-qubit costs excluded and relay/SWAP Hadamards included. On the same fabric, BRIDGE-F matches or exceeds SABRE-based routing on 17 circuits. Rescoring nominal plans with relay-module error equal to direct-CZ error preserves the advantage over ZAP across all 21 circuits, although recompilation can underperform on QFT. These results identify stationary buffer-mediated connectivity as a promising alternative to repeated atom transport under the stated physical assumptions.

quant-ph · cs.AR

版本更新 · v2arXiv:2605.15843

WorldAct: Activating Monolithic 3D Worlds into Interactive-Ready Object-Centric Scenes

Jichen Hu, Jiawei Guo, Jiazhong Cen, Chen Yang, Sikuang Li, Wei Shen

原文摘要

Recent 3D world modeling systems based on generative scene synthesis, such as Marble, can create coherent and explorable 3D environments, yet their outputs are typically static monolithic assets with limited editability and physical interaction. This restricts their use in immersive content creation and embodied simulation, where generated worlds must be actively modified and manipulated. To tackle this challenge, we present WorldAct, a framework that converts static generated 3D worlds into editable and interaction-ready scenes. WorldAct uses a multimodal agent to guide scene decomposition, identify actionable objects, reconstruct geometrically aligned object-level meshes for interaction, and restore the residual background via 3D inpainting. The resulting scenes support object-level editing, collision-aware manipulation, and embodied task execution while preserving global scene coherence. Experiments show that WorldAct enables richer interaction scenarios than the original generated scenes, suggesting a practical path toward editable and interactive 3D world models. Project page: https://sjtu-deepvisionlab.github.io/WorldAct

cs.CV

版本更新 · v3arXiv:2609.37157

From Learner Behavior to Reusable Skills for Effective and Efficient Learner Simulation

Zijian Chen, Zheng Zhang, Miao Jia, Xingchen Hu, Weibo Gao, Linan Yue

原文摘要

Learner simulation aims to reproduce how a particular learner behaves on new tasks. Although Large Language Models (LLMs) can generate increasingly fine-grained learning behaviors, existing approaches often need to repeatedly process a growing interaction history to reconstruct the learner. This introduces additional context and inference costs and makes the acquired learner-specific simulation capability difficult to reuse across different LLMs. We therefore propose Learner2Skill, which externalizes the simulation capability acquired from historical interactions into a persistent and reusable Simulation Skill. The Skill captures the learner's current learning state and recurring response patterns, evolves as new real interactions arrive, and can be adapted to a new LLM through lightweight executor calibration without reconstructing the learner from scratch. Experiments show that Learner2Skill more faithfully reproduces fine-grained learner behavior while reducing overall token cost, and that the same constructed Skills can be effectively reused across different LLM executors.

cs.AI

新论文 · v1arXiv:2611.06684

The Online $k$-Taxi Problem Is Competitive on All Metrics

Christian Coester, Yichen Huang, Tze-Yang Poon

原文摘要

The online $k$-taxi problem is a natural generalization of the $k$-server problem: $k$ taxis must serve a sequence of requests in a metric space. Each request is a pair of points $(r,s)$, and the algorithm serves the request by first moving a taxi to $r$ (to pick up a passenger) and then to $s$ (to drop off the passenger). The cost is the total distance traveled by the taxis without a passenger. Previous bounds for general $k$ all depended on the number of points $n$ or the aspect ratio $Δ$ of the metric space, both of which can be infinite on general metrics, and competitive algorithms were previously known on general metrics only for $k \le 3$. In this work, we give a $9^{k-1}$-competitive randomized algorithm against adaptive online adversaries and, as a corollary, an $81^{k-1}$-competitive deterministic algorithm for the online $k$-taxi problem.

cs.DS

新论文 · v1arXiv:2611.05777

Sublinear Metric Distortion for Matching

Moses Charikar, Alireza Haqi, Jabari Hastings, Yichen Huang, Prasanna Ramakrishnan

原文摘要

We study metric distortion in one-sided matching. Given $n$ agents and $n$ items in an unknown metric space, our goal is to match agents to items so that the total distance between matched pairs is small. However, instead of direct access to the distances in the metric, we are only given each agent's ranking of the items by distance. What is the smallest worst-case approximation factor (called the distortion) that a mechanism can guarantee in expectation? Despite significant interest over the last decade, the best-known guarantee has been the $O(n)$ distortion of Random Serial Dictatorship since the beginning. We give a matching mechanism with distortion $O(n^{\log_4 3})$, the first sublinear upper bound. The key idea is to use a partial matching derived from a truncated version of either Random Serial Dictatorship or the Probabilistic Serial rule. We then complete this matching using a new, general-purpose cluster completion algorithm that ensures unmatched agents receive nearby items. Along the way, we pin down the optimal distortion of Random Serial Dictatorship and Probabilistic Serial up to logarithmic and constant factors, respectively, and give improved lower bounds for truthful mechanisms.

cs.GT · cs.DM · cs.DS

新论文 · v1arXiv:2611.05669

The Harmonic Online $k$-Server Algorithm Is Poly(k)-Competitive

Yichen Huang

原文摘要

The online $k$-server problem is one of the most fundamental problems in online algorithms. In this problem, $k$ servers are located in a metric space and serve a sequence of requests at points in the metric space by moving a server there. The Harmonic algorithm is a very natural algorithm: it chooses a server with probability inversely proportional to its distance from the request. It is also highly efficient, being memoryless and answering each request in $O(k)$ time. However, its competitive ratio has remained unsettled between the $O(2^k \log k)$ upper bound (Bartal and Grove, JACM 2000) and the $k(k+1)/2$ lower bound (Raghavan and Snir, 1994, IBM J. Res. Dev.). We prove that Harmonic is $2k^4$-competitive on every metric space against adaptive online adversaries. Our analysis uses a potential combining corrected electrical energy with the distances of online servers to their assigned adversary locations.

cs.DS

版本更新 · v2arXiv:2610.24973

A Bulow-Klemperer Theorem for Bundling and Improved Approximation Bounds for Simple Auctions

Yichen Huang

原文摘要

We study simple auctions for one additive bidder whose values for $m$ items are independent. Write $\mathrm{REV}$ for the optimal revenue, $\mathrm{SREV}$ and $\mathrm{BREV}$ for the optimal revenues from separate selling and grand bundling, and $\mathrm{BSPA}_2$ for the grand-bundle second-price revenue with two independent bidders. For independent regular items, we prove $\mathrm{REV} \le \mathrm{BSPA}_2 \le 2\,\mathrm{BREV}$. Thus, one additional bidder lets a reserve-free second-price auction for the grand bundle match the optimal revenue from one bidder, regardless of the number of items. This gives a Bulow-Klemperer theorem for bundling and resolves an open question of Beyhaghi, Cai, Feng, Li, and Weinberg (EC 2025) on whether constant competition suffices for bundling. The factor-$2$ bundling approximation is also tight. For general independent items, we prove $\mathrm{REV} \le \mathrm{SREV} + \frac{π^2}{6}\mathrm{BREV} \le 2.645 \max\{\mathrm{SREV}, \mathrm{BREV}\}$. This improves the previous approximation factors of $6$ from Babaioff, Immorlica, Lucier, and Weinberg (FOCS 2014, JACM 2020) and $5.2$ from Ma and Simchi-Levi (AISTATS 2021). For the lower bound, we construct an instance in which pricing three disjoint bundles earns more than $2.11\max\{\mathrm{SREV},\mathrm{BREV}\}$, improving the previous lower bound of $2$ from Rubinstein (ITCS 2016). Together, these results narrow the tight approximation ratio to the range $[2.11, 2.645]$.

cs.GT · econ.TH

版本更新 · v4arXiv:2607.17607

Optimizing the Preconditioner: A Black-box Online-to-Nonconvex Conversion with Static Regret Minimization Oracles

Haichen Hu, David Simchi-Levi

原文摘要

Stochastic nonconvex optimization is central to training deep networks and LLMs in modern machine learning. We give a black-box reduction from stochastic nonconvex optimization to ordinary static regret minimization in online convex optimization (OCO), addressing the open problem posed by Chen and Hazan (2024). Our reduction maintains a predictable gradient tracker, while a black-box online learner $\mathcal A$ selects a preconditioner that transforms this tracker into the update direction. Given a $β$-smooth function with initial gap at most $M$ and an unbiased gradient oracle with variance bounded by $σ^2$, we bound the expected average squared gradient norm by $O(σ\sqrt{Mβ/T}+\sqrt{Mβ}\mathrm{Reg}_T(\mathcal A)/T+\frac{Mβ}{T})$, where $\mathrm{Reg}_T(\mathcal A)$ is the static regret of $\mathcal A$ against the identity preconditioner. Thus, any OCO oracle with $O(\sqrt T)$ regret recovers the classical $O(T^{-1/2})$ convergence rate. We also prove that the additive variance term cannot be removed: every conversion making a constant number of stochastic-oracle queries per round, whose guarantee is stated through static regret against a comparator that the online learner can play, must contain an additive term of order $σ\sqrt{Mβ/T}$ for sufficiently large $T$. Moreover, for every online learner, our conversion bound is tight up to universal constants. We also extend the framework to nonsmooth nonconvex optimization, still relying only on ordinary static regret, and attain the optimal convergence rate for Goldstein-type stationarity. Finally, we conduct numerical experiments on nonconvex objectives to illustrate how the reduction exploits online-selected preconditioners while using the same stochastic-oracle budget as stochastic gradient descent.

cs.LG · math.OC · stat.ML

版本更新 · v3arXiv:2610.39937

BiHashFormer: Hash-Driven Dual-Branch Transformer for Efficient Object Detection in HRW Shots

Jingchen Huang, Wenxi Li, Chenyang Lyu, Moran Liu, Haozhe Lin, Yuchen Guo

原文摘要

Recent advances in gigapixel-level imaging have brought High-Resolution Wide shots to the forefront of research. However, these images present significant challenges: extreme sparsity of foreground, gigapixel-level resolutions and interleaving of foreground and background. This causes traditional detectors and attention mechanisms to be hindered by background, resulting in inefficiency and inaccuracy. To tackle this problem, we propose BiHashFormer, a dual sparsified branch transformer built on hashing. RoI Selector will predict the foreground proportion for top-k selecting target-containing windows. The compression branch processes windows selected and refines attention into HashAttention by discarding half of the key-value pairs, reducing cost and improving efficiency. The compensation branch uses HashMiner to perform low-cost hash searches in the remaining regions, recovering features of missed objects. Experiments confirm that the selection relationship in HashAttention is one-way and reveal the preference of different queries for keys. In experiments on the gigapixel benchmark PANDA, BiHashFormer reduces 37.5% backbone FLOPs while improving AP50 to 81.0%. Moreover, this parallel dual-branch structure allows the model to avoid inference delays caused by inter-branch dependencies, while achieving much greater stability than single-branch models, specially with low rates of RoI selection.

cs.CV

新论文 · v1arXiv:2611.02306

Lie Routing Vision Language Models for View-Robust 3D Understanding

Jingjun Yi, Chen Hu, Qi Bi, Hao Zheng, Huimin Huang, Yongkang, Wei Ji, Yawen Huang 等 10 位作者

原文摘要

Multimodal Large Language Models (MLLMs) have made substantial progress in 3D scene understanding from multi-view RGB observations. Yet a fixed object and query can elicit inconsistent target-related responses across camera viewpoints, a phenomenon we characterize as viewpoint-induced semantic drift. We introduce LieVLM, a Lie routing framework for view-robust 3D understanding. Motivated by the rotational component of viewpoint change, Lie routing adapts visual representations by applying local $\mathrm{SO}(3)$ transformations to projected features within learned three-dimensional subspaces. Geometry tokens extracted by VGGT predict local rotation parameters and channel-wise routing coefficients. A closed-form Padé--Cayley map transforms projected visual features, whose identity-relative corrections are projected back and gated into the original visual and geometry features. This design retains the pretrained semantic and direct spatial context while learning view-dependent feature adjustments. Experiments on 3D captioning, grounding, detection, and spatial reasoning show consistent improvements over additive fusion across training scales and backbones. Controlled ablations examine the roles of geometric conditioning and transformation structure. Viewpoint analyses further reveal systematic variation in the predicted routing states alongside more localized target responses and stronger task performance across viewpoint changes.

cs.CV · cs.LG

新论文 · v1arXiv:2611.01704

MatMoE: Matryoshka Mixture-of-Experts with Dynamic Mixed-Precision Quantization for Efficient Inference

Renjie Wei, Haochen Huang, Chuyu Qiu, Jinqi Wen, Dehao Xu, Meng Li

原文摘要

Mixture-of-Experts (MoE) architecture has been widely adopted in large language models (LLMs) since it reduces the computational cost through sparse expert activation. However, the massive parameter scale of MoE models hinders their deployment on resource-constrained edge devices. Although existing mixed-precision quantization reduces the model size, they are inherently static, fixing the bit-width of each expert across decoding steps. Such static schemes yield suboptimal performance as they overlook the dynamic variation in expert importance during decoding. Furthermore, they fail to adapt to the dynamic latency constraint driven by the fluctuating system utilization. To address these challenges, we propose MatMoE, a dynamic mixed-precision quantization framework for efficient MoE inference. MatMoE features: (1) a Matryoshka MoE PTQ algorithm that enables experts to operate flexibly at multiple precisions within a stringent memory budget; (2) a floor-and-compensation quantization method that eliminates redundant memory accesses in the Matryoshka representation; and (3) a dynamic bit-width allocation strategy that determines the precision for each expert based on their routing scores to satisfy the dynamic latency requirement. We implement our MatMoE system atop llama.cpp for both CPU and GPU platforms. Extensive evaluations across three modern MoE models and challenging benchmarks demonstrate that MatMoE achieves $4.0\% \sim 18.3\%$ higher accuracy than state-of-the-art static mixed-precision method. Compared to llama.cpp, MatMoE delivers $1.3\times \sim 3.7\times$ speedup on Raspberry Pi 5, and $1.2\times \sim 1.3\times$ speedup on NVIDIA RTX 4090 and Jetson AGX Orin.

cs.LG · cs.AI

新论文 · v1arXiv:2610.40293

MAServeBench:面向分层协作工作负载的多智能体LLM服务系统基准测试

MAServeBench: Benchmarking Multi-Agent LLM Serving Systems with Hierarchical Collaboration Workloads

Shengxuan Qiu, Haochen Huang, Shuzhang Zhong, Zizhuo Fu, Zebin Yang, Jianming Tong, Meng Li

AI 速读 基于摘要

提出一个把多智能体协作规则作为可配置维度、用于评测LLM服务系统与GPU/NPU硬件行为的基准。

cs.MA

查看原文摘要

Large language model (LLM)-based multi-agent systems (MAS) coordinate specialized agents to solve complex tasks, turning each user request into interdependent model and tool operations. Unlike independent requests, these operations become ready at different times and consume evolving context as agents coordinate and exchange results. This creates changing patterns of concurrency, computation, and memory demand that can expose different bottlenecks in LLM serving hardware. Yet benchmarking these effects requires capturing the collaboration rules that generate serving workloads, beyond the requests observed during execution. We present MAServeBench, a benchmark that treats multi-agent collaboration as a configurable dimension of serving workloads for LLM serving and hardware evaluation. Its composable stages capture how agents coordinate and exchange context in documented MAS workflows, enabling native executions to generate different workloads even when task inputs and tools remain unchanged. Hierarchical Semantic Traces preserve realized operations and dependencies for reproducible comparisons of legal release constraints at fixed work. Task-linked measurements connect these workloads to serving execution and GPU/NPU hardware behavior, revealing why serving efficiency does not always translate into timely task completion. The benchmark code is available at https://github.com/ShengxuanQiu/MAServeBench.

新论文 · v1arXiv:2610.39946

ElasticFormer: Detecting Objects in HRW Shots via Elastic Computing Vision Transformer

Wenxi Li, Jingchen Huang, Chenyang Lyu, Moran Liu, Haozhe Lin, Guiguang Ding, Yuchen Guo

原文摘要

Recent advances in gigapixel-level imaging have brought High-Resolution Wide shots to the forefront of research. However, these images present significant challenges: extreme sparsity of foreground, gigapixel-level resolutions and diverse target counts. This makes traditional close-up detectors inaccurate and slow as they are overwhelmed by the background. Although previous research has explored sparse backbones, their fixed sparsity patterns lack the adaptability required to handle diverse target numbers. To address this, we introduce ElasticFormer, a sparse backbone that dynamically allocates computational resources based on foreground proportion. After scoring windows based on variance, proposed ElasticSelector module will predict the foreground proportion for top-k selection. The mechanism guides the model to select target-containing windows, scaling resources in areas where objects are clustered. We introduce a novel loss function combined with the 3-phase training strategy for ElasticSelector, allowing it to function properly when bounding box annotations are missing. A WSOD study is carried on PASCAL VOC 2007 to evaluate its extensibility. Further, ElasticNet is created to verify its backbone-agnostic nature. In experiments on the PANDA gigapixel benchmark, ElasticFormer reduces backbone FLOPs by 80% while achieving a significant improvement in AP50 when compared to fixed-ratio sparse methods.

cs.CV

新论文 · v1arXiv:2610.39782

Assessing AI Companion Chatbot Safety for Vulnerable Users: A Crisis-Stage Multi-Platform Evaluation

Huanyue Liao, Seok Min Lim, Yan Ling Lim, Shi Min Cheng, Kellin Pelrine, Chen Hui Ong

原文摘要

AI companion chatbots are increasingly used by vulnerable individuals, including adolescents in suicidal crisis, for emotional support. Existing safety evaluations focus on base models, leaving gaps in product-level assessment, crisis-stage sensitivity, and comparisons between companion apps and general-purpose LLMs. We introduce MIRROR-Eval (Multi-turn Interaction Risk Review: Observing suicidal-crisis Responses), a dataset of 960 annotated 20-turn conversations across eight deployed platform configurations, six clinically grounded personas, and two crisis stages: ideation and explicit planning. An LLM-as-a-Judge pipeline applies an expert-reviewed rubric covering user risk severity, chatbot risk awareness, and intervention behavior. We observe a three-tier harm structure: companion apps (Character.AI and Replika) and Grok 4.3 form a high-harm tier; Gemini 3.1 Pro is mid-range; GPT-5.5 and Claude Sonnet 5 show substantially lower harm. At planning, companion apps show increased glorification and normalisation, advancement toward harmful action, and harmful methods and instructions. Crisis recognition and empathetic language do not ensure safe responses. These findings highlight the need for product-level safety evaluation across crisis stages. Dataset and code: https://huggingface.co/datasets/imda-biztech/MIRROR-Eval. Warning: this paper includes examples that may be offensive or harmful.

cs.HC

版本更新 · v2arXiv:2609.17474

Iterative Calibration and Learning: Mitigating Teacher Bias in LLM Distillation without Target-Domain Reward Feedback

Haichen Hu, Yuheng Zhang, David Simchi-Levi

原文摘要

Large language model (LLM) distillation can propagate teacher bias under covariate shift when target reward feedback is unavailable. We propose Iterative Calibration and Learning (IterCaL), which alternates source-reward teacher calibration through token-level branching with student learning on target prompts. We prove polynomial convergence in expected average KL to the oracle student, the reference-regularized target-reward maximizer within a potentially misspecified student class; a quantified source-to-target calibration discrepancy gives an explicit additive true-return term. Our separation shows that regularized direct teacher matching retains positive oracle-KL error at its population optimum for every shared reference penalty, even when the teacher is stronger than every student. On the same instance, IterCaL converges to the oracle using source rewards alone.

cs.LG · cs.AI · math.ST · stat.ML

新论文 · v1arXiv:2610.36616

Normalized Cuts and Data Clustering via Matrix Fractional Programming

Yannan Chen, Beichen Huang, Licheng Zhao, Chris Ding, Kaiming Shen

原文摘要

The normalized cut (NCut) problem is a fundamental one in the area of data clustering. Despite its fractional structure, the NCut problem is seldom considered from a fractional programming (FP) perspective in the existing literature. The main reason is that the conventional FP method is inadequate for this problem case: the classical Dinkelbach's transform can only handle a single ratio and hence is limited to the two-class clustering, whereas the multi-class clustering involves multiple ratio terms. This work proposes a novel so-called multidimensional quadratic transform that is capable of dealing with the multi-ratio case; it recasts the NCut problem into a bipartite matching problem---which can be immediately solved by linear search. Furthermore, we explore the connection between the proposed FP method and the minorization-maximization (MM) theory, and in turn characterize the convergence behavior of our algorithm. Finally, we conduct experiments to numerically demonstrate the advantages of the proposed FP-based algorithm in a variety of data clustering and image segmentation tasks.

cs.LG · eess.SP

版本更新 · v10arXiv:2507.16199

LLM Abstention Can Be a Prompt Artifact, in Addition to Genuine Uncertainty

Zipeng Ling, Shuliang Liu, Yuehao Tang, Junqi Yang, Shenghong Fu, Seonil Son, Chen Huang, Kejia Huang 等 11 位作者

原文摘要

Large Language Models (LLMs) are increasingly trained to abstain from answering questions they are unsure about. However, this ability is often misapplied: in real-world applications, user prompts sometimes contain elements of uncertainty, which lead LLMs to abstain even on problems they are capable of solving. We argue that LLM abstention is not only an expression of genuine uncertainty; it can also be an artifact largely shaped by prompts. We name this phenomenon *Abstention Inflation*. We add "Unknown" as an extra option for LLMs to choose from; experiments show serious accuracy drops on True/False Questions (TFQs). Replacing "Unknown" with an unrelated random word produces a similar effect. We argue that LLMs are trained to imitate the surface pattern of abstention, rather than to express genuine uncertainty. Based on ten experimental settings, we support four claims that form a progressive argument: **(C1)** *Abstention Inflation* can be triggered by the presence of an extra option, not by genuine uncertainty; **(C2)** it makes the models deny they can answer, even when they can; **(C3)** it is a later-layer output override, as the reasoning traces and mid-layer representations preserve correct answers; **(C4)** it is not stochastic noise: it results from various factors, emerges through instruction tuning, is boosted by problems' higher difficulty, and can be mitigated at larger model sizes.

cs.CL

新论文 · v1arXiv:2610.35773

A Unified Arena for LLM Agents across Diverse Eurogame

Fengrui Zhang, Xiang Li, Ruomeng Ding, Jiayi Yuan, Yuchen Huang, Zhaoyang Han, Xiaoliang Wang, Daoyuan Chen 等 9 位作者

原文摘要

Interactive environments offer a practical way to evaluate large language model (LLM) agents, and competitive games extend this evaluation to decisions shaped by competing goals and other agents' choices. Yet existing evaluations confined to individual games or execution frameworks leave unclear which competitive advantages persist across settings. We introduce DECE (Diverse Eurogame Competitive Environments), a unified arena spanning ten Eurogames with shared interfaces and evaluation protocols. These games combine lasting strategic commitments, competition over shared resources, and multiple routes to victory. Our primary study evaluates 12 model-harness configurations, spanning nine models and four execution harnesses, in 480 three-player matches across ten games. The main and supplementary studies comprise 1,032 completed matches in total. These comparisons reveal broad cross-game ranking agreement alongside game-specific strengths. Matched studies show that the same model can perform differently across execution harnesses, adding another source of variation to these profiles. These performance advantages do not consistently translate into better token or cost efficiency, making resource use an additional consideration when comparing configurations. Together, the findings support evaluating complete agent configurations by both their competitive performance across games and the resources needed to achieve it.

cs.AI

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1158ACL Anthology · ACL 2026

EyeMulator: Improving Code Language Models by Mimicking Human Visual Attention

Yifan Zhang SalesForce.com, Chen Huang National University of Singapore, Yueke Zhang, Jiahao Zhang Vanderbilt University, Toby Jia-Jun Li University of Notre Dame, Collin McMillan University of Notre Dame, Kevin Leach Vanderbilt University, Yu Huang Vanderbilt University

原文摘要

Code Language Models (CodeLLMs) traditionally learn attention based solely on statistical input-output token correlations (“machine attention”). In contrast, human developers rely on intuition, selectively fixating on semantically salient tokens during program comprehension. We present EyeMulator, a model-agnostic technique to align CodeLLM attention with human visual attention without architectural changes. By extracting scan paths from eye-tracking data, we derive token-level attention weights used to augment the loss function during fine-tuning. This induces the model to mimic human focus. Our evaluation across StarCoder, Llama-3.2, and DeepSeek-Coder shows that EyeMulator significantly outperforms baselines, achieving gains of over 30 CodeBLEU points in translation and up to 22 BERTScore points in summarization. Ablation studies confirm that these gains stem directly from replicating human attention dynamics. Artifacts are available at https://zenodo.org/records/17205682.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1428ACL Anthology · ACL 2026

SpeechMedAssist: Efficiently and Effectively Adapting Speech Language Models for Medical Consultation

Sirry Chen, Jieyi Wang, Wei Chen Huazhong University of Science and Technology, Zhongyu Wei Fudan University

原文摘要

Medical consultations are intrinsically speech-centric. However, most prior works focus on long-text-based interactions, which are cumbersome and patient-unfriendly. Recent advances in speech language models (SpeechLMs) have enabled more natural speech-based interaction, yet the scarcity of medical speech data and the inefficiency of directly fine-tuning on speech data jointly hinder the adoption of SpeechLMs in medical consultation. In this paper, we propose SpeechMedAssist, a SpeechLM natively capable of conducting speech-based multi-turn interactions with patients. By exploiting the architectural properties of SpeechLMs, we decouple the conventional one-stage training into a two-stage paradigm consisting of (1) Knowledge Capability Injection via Text and (2) Modality Re-alignment with Limited Speech Data, thereby reducing the requirement for medical speech data to only 10k synthesized samples. To evaluate SpeechLMs for medical consultation scenarios, we design a benchmark comprising both single-turn question answering and multi-turn simulated interactions. Experimental results show that our model outperforms all baselines in both effectiveness and robustness in most evaluation settings.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1618ACL Anthology · ACL 2026

E2EDev: Benchmarking Large Language Models in End-to-End Software Development Task

Jingyao Liu, Chen Huang National University of Singapore, Zhizhao Guan, Wenqiang Lei Sichuan University, Yang Deng Singapore Management University

原文摘要

The rapid advancement in large language models (LLMs) has demonstrated significant potential in End-to-End Software Development (E2ESD). However, existing E2ESD benchmarks are limited by coarse-grained requirement specifications and unreliable evaluation protocols, hindering a true understanding of current framework capabilities. To address these limitations, we present E2EDev, a novel benchmark grounded in the principles of Behavior-Driven Development (BDD) to assess whether the generated software meets user needs through mimicking real user interactions. E2EDev comprises (i) a fine-grained set of user requirements for each target software project (ii) multiple BDD test scenarios with corresponding Python step implementations for each requirement, and (iii) a fully automated testing pipeline built on the Behave framework. By evaluating various E2ESD frameworks and LLM backbones with E2EDev, our analysis reveals a persistent struggle to effectively solve these tasks, underscoring the critical need for more effective and cost-efficient E2ESD solutions. Our codebase and benchmark are available at https://github.com/SCUNLP/E2EDev.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1668ACL Anthology · ACL 2026

METER: Evaluating Multi-Level Contextual Causal Reasoning in Large Language Models

Pengfeng Li, Chen Huang National University of Singapore, Chaoqun Hao, Hongyao Chen, Xiao-Yong Wei, Wenqiang Lei Sichuan University, See-Kiong Ng National University of Singapore

原文摘要

Contextual causal reasoning is a critical yet challenging capability for Large Language Models (LLMs). Existing benchmarks, however, often evaluate this skill in fragmented settings, failing to ensure context consistency or cover the full causal hierarchy. To address this, we pioneer METER to systematically benchmark LLMs across all three levels of the causal ladder under a unified context setting. Our extensive evaluation of various LLMs reveals a significant decline in proficiency as tasks ascend the causal hierarchy. To diagnose this degradation, we conduct a deep mechanistic analysis via both error pattern identification and internal information flow tracing. Our analysis reveals two primary failure modes: (1) LLMs are susceptible to distraction by causally irrelevant but factually correct information at lower level of causality; and (2) as tasks ascend the causal hierarchy, faithfulness to the provided context degrades, leading to a reduced performance. We belive our work advances our understanding of the mechanisms behind LLM contextual causal reasoning and establishes a critical foundation for future research. Our code and dataset are available at https://github.com/SCUNLP/METER.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1877ACL Anthology · ACL 2026

BOSCH: Black-Box Binary Optimization for Short-Context Attention-Head Selection in LLMs

Abbas Ghaddar Huawei Technologies Canada Co., Ltd., Ivan Kobyzev Noah's Ark Lab, Huawei Technologies Canada Co., Boxing Chen Huawei Technologies Ltd., Yufei Cui

原文摘要

Post-training hybridization of large language models (LLMs) often replaces quadratic self-attention with sliding-window attention (SWA) to reduce KV cache usage and improve latency. Existing hybridization schemes are typically defined either at the layer level (e.g., interleaving) or at the head level via static rankings from local to global. Layer-level schemes ignore that local and global dependencies are routed through heads within the same layer, while static head-level rankings suffer from entanglement: a head’s local/global behavior can change after hybridization. We propose BOSCH, Black-box Binary Optimization for Short-context Head Selection, a training-free method that formulates the problem as a Large Neighborhood Search and decomposes it into three subproblems: (i) layer-importance detection via small-budget black-box probes, (ii) adaptive per-layer SWA-ratio assignment based on these sensitivities, and (iii) grouped head-level optimization within ratio buckets. Extensive experiments on 4 LLMs ranging from 1.7B to 30B parameters, across 4 SWA ratios, show that BOSCH consistently outperforms layer-level heuristics and 6 strong static head-level methods, with larger gains at higher SWA ratios. Under continual pretraining, BOSCH recover original long-context performance faster and to a higher level. Analysis of the selected heads reveals substantial turnover for BOSCH across different SWA ratios, underscoring the importance of performing head-level selection for each target ratio rather than relying on fixed locality rankings.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表