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

+1112新增论文
+875其中 AI 相关
+133已有论文修订
+97新增 AI 导读

arXiv 最近完成扫描:2026-11-20 04:39 · 全站最近入库:2026-11-19 19:09各信源进度 →

统计口径与数据新鲜度

统计 2026-11-19 05:31 — 2026-11-20 05:31(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:2026-11-20 04:39最近发现新增版本:2026-11-19 19:09库内论文最新更新时间:2026-11-19 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。

27,982 已收录论文23,249 AI 相关8,559 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 33 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-20 04:38(北京时间)
版本更新 · v2arXiv:2605.10093

RFAmpDesigner: A Two-Level LLM-guided Optimization Framework for Automated Differential CMOS RF Amplifier Sizing

Hang Lu, Guochang Li, Qianyu Chen, Huiyan Gao, Shaogang Wang, Xuanyu He, Yiwei Liu, Gaopeng Chen 等 12 位作者

原文摘要

Automated RF amplifier sizing is challenging because the design process is inherently hierarchical, the parameter space is high-dimensional and strongly coupled, and high-fidelity circuit simulation is expensive. Conventional optimization and reinforcement learning methods typically require many iterations and make limited use of RF domain knowledge, making them costly under tight simulation budgets. Large language models (LLMs) provide broad prior knowledge and reasoning capabilities, but are unreliable for direct numerical optimization. This paper presents RFAmpDesigner, a two-level optimization framework for automated schematic-level sizing of differential CMOS RF amplifiers. At the outer level, the circuit-parameter space is abstracted into a resource-allocation space, where LLMs make high-level allocation decisions and adjust them according to feedback from the inner level. At the inner level, RF design knowledge is used to decompose the flat sizing problem into a sequence of low-dimensional subproblems. Conventional optimizers then solve these subproblems to find circuit implementations consistent with the assigned resources. This structure allows numerical optimizers to focus on low-dimensional searches in the circuit-parameter space, while the LLM reasons and acts over a physically meaningful design space using simulation feedback. The main evaluation covers low-noise amplifier (LNA), driver amplifier (DA), and power amplifier (PA) designs within the target differential CMOS RF amplifier family. Experimental results show that it achieves a higher success rate under a limited simulation budget in both easy and hard regimes.

cs.AR

版本更新 · v2arXiv:2611.21072

Sage S2.1 Flash Technical Report

Kun Cai, Haodong Chen, Guangran Cheng, Qingxiu Cheng, Xu Dai, Guanting Dong, Qiming Ge, Lixin Gu 等 206 位作者

原文摘要

We introduce Sage S2.1 Flash, a 320-billion-parameter general-purpose mixture-of-experts model designed to reason over scientific evidence and act in interactive environments. Sage Sci-Perception uses dedicated encoders for molecular graphs and biological sequences and represents spectra as ordered patches of measurements. Modality-specific projectors map these representations into the input space of a shared backbone, which processes them alongside text and images. To develop the model's agent capabilities, we construct verifiable environments for tool use, search, and coding, as well as specialized workflows involving machine learning experiments and scientific software. Task completion is checked against execution results, changes in environment state, or submitted artifacts. We evaluate Sage S2.1 Flash on 33 benchmarks across seven domains. In the primary comparison against five baselines, it ranks first on OfficeQA Pro (72.2), MLE-bench Lite (90.3 HumanRank), Harvey LAB, and SpreadsheetBench 2, and second on OSWorld-Verified (85.2), WebArena-Verified (67.6), and DeepResearch Bench II (54.2). These results show particular strengths in using documentary evidence, producing verifiable workbooks, and carrying out machine learning tasks, alongside competitive performance in desktop and browser workflows that require the model to act on visual feedback.

cs.AI

新论文 · v1arXiv:2611.22350

An AI-Driven Assistant for Human-Centric Building Digital Twin Systems: Design, Implementation, and Comparative User Evaluation

Leonardo Messi, Erika Anneli Pärn, Juan Diego Castaño Molina, Oleksandr Anyshchenko, Dijar Selami, Peihang Luo, Borja García de Soto

原文摘要

Smart buildings generate streams of operational data, yet this richness rarely translates into accessible, decision-ready information for stakeholders who inhabit and manage them. Current Building Digital Twin Systems (BDTS) often remain oriented toward technical specialists and rely on static, one-size-fits-all interfaces. This paper presents the design, implementation, and comparative user evaluation of an AI-driven conversational assistant integrated into a human-centric BDTS for Indoor Environmental Quality monitoring, combined with adaptive information presentation and auto-suggest functionality. The prototype was deployed at New York University Abu Dhabi. A controlled, two-arm, between-subjects, quasi-experimental study with wave-based, non-randomized assignment compared the Baseline interface with the Experimental interface. Results showed significant reductions in task-completion time and perceived workload, while accuracy and usability improved but did not meet the adjusted significance criterion. The study contributes an implemented and empirically evaluated architecture that separates generative interaction from deterministic environmental assessment and coordinates conversational responses with adaptive visual presentation.

cs.HC · cs.AI

版本更新 · v5arXiv:2610.18826

WorldOracle: Benchmarking Action-Conditioned World Models against Real-World Executions

Haoyi Duan, Liliang Chen, Nan Wang, Yang Zhang, Yuhang Lu, Xiaotong Chen, Wei-Cheng Tseng, Shenyuan Gao 等 16 位作者

原文摘要

We introduce WorldOracle, a benchmark specifically designed to evaluate action-conditioned world models (ACWMs) against paired real-world executions. Given the same initial observation and low-level action sequence, WorldOracle asks whether a predicted rollout reproduces what happens in the real world, including physical consequences and task-state transitions. We evaluate models along four complementary dimensions: physical state correctness, interaction spatial alignment, key interaction area fidelity, and multi-view action coherence. To support this evaluation, we curate real-world training and evaluation suites that cover different interaction dynamics and embodiments. The evaluation suite contains long, multi-segment episodes with both successful and failed executions, enabling evaluation across varied action outcomes and behavior distributions. Experiments on seven representative ACWMs reveal distinct model strengths and failure modes. WorldOracle provides a systematic foundation for measuring and advancing action-conditioned world modeling for manipulation. We release the datasets, evaluation code, and leaderboard: https://worldoracle-benchmark.github.io/.

cs.RO · cs.CV

新论文 · v1arXiv:2611.18262

Exact-Quota Facility Location: A Transfer Framework and Three Approximation Guarantees

Hang Luo

原文摘要

Facility location with outliers asks which facilities to open and which clients to serve, subject to a prescribed service quota. We give a framework for transferring facility-location cost certificates to this exact-quota setting. The framework yields deterministic polynomial-time approximation guarantees of $571/376<1.52$ for metric facility location with outliers, strictly less than $1.853$ for priority facility location with outliers, and $2.16995<2.17$ for squared-metric facility location with outliers. Each algorithm serves the prescribed number of clients and pays the original facility-opening costs. The framework separates initialization from improvement. A growth process is stopped when it first reaches the quota; deleting excess clients from its last batch leaves a payment deficit supported on a bounded number of facilities. Greedy additions then reoptimize the served client set. A threshold-coverage representation makes the resulting service cost supermodular and allows the usual density analysis to compare different client sets. Finally, enumerating and prepaying a constant number of expensive facilities absorbs the exceptional costs. A transfer theorem states exactly which initialization certificate is sufficient and how many prepaid facilities it requires. The three applications supply this certificate in different ways. The metric case uses a quota prefix of reassigning JMS; the priority case uses persistent contributions and a cleanup ordered by service level; and the squared-metric case uses a local factor-revealing certificate valid for arbitrary client subsets of the same run. We provide complete finite proofs of the numerical interfaces, including an exact rational certificate for the squared-metric bound.

cs.DS

新论文 · v1arXiv:2611.17845

Tight Quantum Query Bounds for Minimum Cut and Multicut on Dense Graphs

Tianhang Lu, Runtian Ren, Shengcai Liu

原文摘要

We determine the dense worst-case quantum query complexity of exact minimum multicut on simple unweighted undirected graphs. With at most $k$ explicit demand pairs, our adjacency-array algorithm uses $\OO(n+\min\{m,n^{3/2}\sqrt{k}\})$ queries; a matching lower bound holds for every $1\le k\le\lfloor n/2\rfloor$ and sufficiently large $n$, with constants uniform in $k$. In particular, the $k=1$ case gives an $\OO(n^{3/2})$ query algorithm for prescribed $s$--$t$ cut. The upper bound combines degree-sensitive spectral estimates of local improvement with refinement by unions of label classes. These steps retain every original optimum and permit exact acquisition without an additional label-count factor. Separately, we prove an $Ω(n^{3/2})$ global minimum-cut lower bound on dense connected graphs, matching the existing upper bound up to polylogarithmic factors. It holds for the value and for any optimum shore, even with known degrees and sorted neighbor arrays; regular hard instances have balanced optimum cuts. Both lower bounds use degree-preserving switches and classify optimum outputs across a hidden threshold. Our upper bound concerns input queries, not total running time: computation on acquired data may be exponential, even in the single-demand case. Matching bounds at each prescribed sparse edge count are not established.

cs.DS · cs.CC

版本更新 · v2arXiv:2610.02771

Nearly Optimal Fixed-Confidence Best-Arm Identification with 1-Bit Feedback

Khang Luong, Dinh Thai Son, Hoang Ta, Hung The Tran, Tuan Quang Dam

原文摘要

We study fixed-confidence best-arm identification under strict 1-bit feedback constraints. At each round, the learner selects an arm and a query set, and receives only a single bit indicating whether the sampled reward belongs to that set. We consider a distribution-free finite-variance setting with arm-wise localization, where direct empirical mean estimation is no longer available and clipping becomes unavoidable. We first formulate a time-uniform 1-bit mean-estimation primitive based on randomized threshold queries and a clipped tail-integral identity. We then embed this primitive into candidate-challenger best-arm identification algorithms. A fixed-clipping algorithm gives a simple anytime $(ε,δ)$-PAC guarantee, while a phased adaptive-clipping algorithm matches the clipping level to the current resolution and yields a gap-adaptive sample complexity. We also prove a $K$-arm worst-case information-theoretic lower bound showing that the logarithmic penalty caused by finite-variance 1-bit feedback is intrinsic. This bound matches the leading dependence of the phased algorithm up to logarithmic factors.

cs.LG · cs.AI · stat.ML

版本更新 · v2arXiv:2611.12799

Agentic-OPD: Recovering Forgotten Agent Capabilities via On-Policy Distillation for Agentic Tasks

Tingcheng Bian, Kangye Gong, Jinchang Luo, MingQuan Cheng, Feiran Liu, Ni Li, Xiaoling Xia, Haiwei Wang 等 9 位作者

原文摘要

Reinforcement learning is now the mainstream paradigm for training LLMs on agentic tasks, yet standard practice keeps only the best-validation checkpoint and discards the rest. We show that this is lossy: in GRPO training of LLM agents, the best-validation checkpoint fails on problems that earlier checkpoints from the same run had already solved, and the union of problems solved across checkpoints far exceeds any single one. The discarded checkpoints thus form a teacher pool obtained at no additional training cost, which on-policy distillation (OPD) can mine to recover the lost coverage. Mining this pool raises two challenges: i) which checkpoint to distill from, as no checkpoint dominates the others and complementarity to the student is problem-specific; ii) where to trust it, as even a well-chosen teacher is only locally superior, so indiscriminate distillation overwrites behavior the student has already mastered. We propose Agentic-OPD, which treats the best-validation checkpoint as the student and all saved checkpoints as frozen teacher candidates, using the student's evolving capability gaps to guide both teacher selection and supervision filtering: the HCN policy (Highest Coverage Next) dynamically selects teachers by weighted coverage of the student's unsolved problems, while the LAD module (Locality-Aware Distillation) confines transfer to failed student rollouts on problems where the teacher performs better. Across five agentic task domains and 24 benchmarks and task subsets, Agentic-OPD improves over its initialization checkpoint by up to 7.4 points, with a 4B student surpassing the latest frontier models GPT-5.6 and DeepSeek-V4.1 on deep search. Agentic-OPD recasts a finished RL run as a reusable pool of policies and establishes post-RL capability recovery as a stage orthogonal to improvements in the RL algorithm itself.

cs.LG · cs.AI

版本更新 · v3arXiv:2606.03303

LEAP:用智能体框架为形式数学增强大语言模型

LEAP: Supercharging LLMs for Formal Mathematics with Agentic Frameworks

Po-Nien Kung, Linfeng Song, Dawsen Hwang, Jinsung Yoon, Chun-Liang Li, Simone Severini, Mirek Olšák, Edward Lockhart 等 13 位作者

AI 速读 基于摘要

LEAP 是一个智能体框架,让通用基础模型通过与 Lean 编译器持续交互,把非形式化推理转化为可机械验证的形式化证明,并在 2025 年 Putnam 竞赛与自建 Lean-IMO-Bench 上取得作者报告的高求解率。

cs.AI

查看原文摘要

Large Language Models (LLMs) exhibit strong informal mathematical reasoning but struggle to generate mechanically verifiable proofs in formal languages like Lean. We present LEAP, an agentic framework that enables general-purpose foundation models to achieve state-of-the-art performance on automated formal theorem proving. LEAP leverages foundation model capabilities, such as informal reasoning, instruction following, and iterative self-refinement. By decomposing complex problems into smaller units, the system bridges formal proof construction with informal blueprints through continuous interaction with the Lean compiler. To provide a rigorous evaluation beyond increasingly saturated benchmarks, we introduce Lean-IMO-Bench, a benchmark of IMO-style problems formalized in Lean, with short statements yet highly non-routine and multi-step proofs across a wide range of difficulty levels. Empirically, on the latest 2025 Putnam Competition, an annual mathematics competition for undergraduate students in North America, LEAP solves all 12 problems, matching recent breakthroughs by frontier formal mathematical models. On Lean-IMO-Bench, LEAP boosts the one-shot formal solve rate of general-purpose LLMs from below 10% to 70%, notably surpassing the 48% benchmark set by a specialized, gold-medal-caliber IMO system. Furthermore, we demonstrate LEAP's research-level utility by autonomously formalizing complex proofs for open combinatorial challenges, including a verified proof for a key subproblem in Knuth's Hamiltonian decomposition of even-order Cayley graphs.

新论文 · v1arXiv:2611.16063

Vocabulary-Free Next-Event Prediction on Temporal Graphs

Zhishang Luo, Venu Madhav Tammali, Amila Weerasinghe, Osso Vahabzadeh, Olcay Boz, Xue Hao, Yusu Wang

原文摘要

Autoregressive transformer training through next-token prediction has driven remarkable advances in language modeling and has become increasingly effective for time-series modeling. Yet its success in temporal graph learning remains limited, and most temporal link-prediction methods are discriminative rather than autoregressive. While temporal graphs can have a natural serialization through chronological ordering edges, they lack an obvious graph-independent tokenization: existing node tokenization often uses a vocabulary that grows with the number of nodes, limiting scalability and cross-graph training. We propose GTGT, Generative Temporal Graph Transformer, a simple yet powerful vocabulary-free framework for self-supervised next-event prediction on temporal graphs. It predicts events for observed destinations, behavioral descriptors for novel destinations, and time gaps before new event. These targets have shared meanings across graphs, allowing one transformer to train across disjoint node sets without node-specific learned parameters. Local node identifiers let attention capture graph structure: Theoretically, we can show that our transformer-based model has the capacity for time-respecting message passing mechanisms. On Temporal Graph Benchmark (TGB), our GTGT not only achieves comparable or better link-prediction performance than established baselines, but also has much lower training cost. These results open the door to temporal graph pretraining for forecasting and other downstream tasks, providing a concrete step toward temporal graph foundation models.

cs.LG

新论文 · v1arXiv:2611.14949

RSI-PDEbench: A Benchmark for Self-Improving Agents in PDE Solving

Binghang Lu, Lang Cao, Changhong Mou, Yige Sun, Cindy Xiangrui Kong, Zheyuan Deng, Tian Yuan, Bing Hu 等 11 位作者

原文摘要

Identifying and solving partial differential equations (PDEs) from sampled solution data often requires substantial expert effort in equation discovery, coefficient estimation, and numerical solver configuration, especially under noise and distribution shifts. Can large language model (LLM) agents automate this workflow accurately and efficiently? To investigate this question, we introduce \bench{}, a benchmark of 1,001 cases spanning 42 equation templates, including both classical and fractional PDEs. It provides three evaluation settings: Full evaluates performance on noise-free solution data; Noisy tests robustness to Gaussian noise with an RMS equal to 50\% of the clean-signal RMS; and Hard-OOD tests challenging out-of-distribution generalization to unseen equation templates and shifted coefficient ranges. We further introduce RSI-PDE, a recursively self-improving agent harness that shortlists candidate equations by residual loss, uses relative $L^2$ feedback to refine coefficients and numerical solvers, and applies a validation gate to policy and memory updates. We compare RSI-PDE with single-run baselines and native agent harnesses (Claude Code or Codex) across four frontier models. Across the three settings, RSI-PDE improves the pass rate averaged across the four models by 106.5--181.9\% relative to single-run baselines and by 3.0--19.6\% relative to native harnesses, while reducing total recorded model calls by 46.9\% across the 12 model--setting pairs. These results highlight the potential of validated recursive self-improvement in LLM agents to advance scientific computing.

cs.AI

版本更新 · v3arXiv:2606.15893

BALTO:面向幻觉缓解的平衡令牌级策略优化

BALTO: Balanced Token-Level Policy Optimization for Hallucination Mitigation

Ning Li, Zixuan Guo, Yan Xu, Wenbo Fei, Yifan Niu, Chang Luo, Yasheng Wang, Weiwen Liu 等 10 位作者

AI 速读 基于摘要

提出 BALTO 框架,通过将声明级事实核验投影为令牌级标签并引入平衡的令牌级信用分配,缓解大语言模型在知识密集型场景中的幻觉问题。

cs.CL

查看原文摘要

Hallucinations remain a major obstacle to deploying large language models (LLMs) in knowledge-intensive settings, where generated responses must be faithfully grounded in provided evidence. Reinforcement learning (RL) is a promising direction for hallucination mitigation, but response-level faithfulness rewards suffer from a granularity mismatch: localized hallucinations can cause supported content to receive spurious penalties. Although recent work introduces fine-grained feedback such as claim-level verification and token-level rewards, unbalanced credit assignment can still induce length, verbosity, or optimization-noise biases. We propose BALTO, a Balanced Token-level Policy Optimization framework for hallucination mitigation. BALTO extracts checkable factual claims, verifies them against the reference context, and projects claim-level judgments to token-level labels. A balanced token-level credit assignment mechanism is introduced into the framework. This design redistributes probability mass from unsupported content toward faithful content, rather than suppressing the entire response. We systematically analyze the limitations of response-level rewards from a theoretical standpoint, and prove BALTO's advantages in training stability and optimization efficiency for hallucination mitigation. Experiments on ConFiQA, RAGTruth, and FinLLM-Eval show that BALTO achieves the highest faithfulness across all six model--benchmark settings and consistently outperforms existing post-training baselines in Q-Score, demonstrating a stronger faithfulness--informativeness trade-off. The code is released at https://github.com/NingLi670/BALTO.

新论文 · v1arXiv:2611.14364

ARE-bench:在真实研究者会话与工作区中评测智能体的研究能力

ARE-bench: Benchmarking Agents' Research Ability in Real-World Researcher Sessions and Workspaces

Jiahang Lin, Wanying Ren, Wanghan Xu, Huayu Sha, Yang Shu, Zhonghang Lu, Chengjun Pan, Yi Dong 等 27 位作者

AI 速读 基于摘要

ARE-bench 用真实研究者的会话与未清理工作区构建 75 个任务,检验研究智能体能否在延续项目时遵守先前决策与工作区状态。

cs.CL

查看原文摘要

Research agents achieve increasingly strong benchmark scores, yet these scores do not necessarily translate into reliable delivery in ongoing research projects. Many research-agent benchmarks emphasize standalone tasks and relatively self-contained workflows in clean or curated workspaces. However, researchers ask agents to complete one step within an ongoing project, where requests depend on earlier session decisions and workspaces contain unfinished files, outdated versions, and failed attempts. We introduce ARE-bench to evaluate whether research agents advance ongoing projects while respecting prior decisions and existing workspace state, grounded in real session context and in-progress workspaces. ARE-bench includes 75 tasks built on agent sessions and full workspaces from recruited researchers' own projects, spanning seven research stages, 33 requirements, and 150 sub-requirements, with 1,173 binary checklist items in total. Each task is anchored at a real user request, preserves its requirements, and summarizes the earlier requests still in effect as background. The complete workspace is replayed to that boundary without cleanup, preserving artifacts from earlier turns. Each checklist is compiled from an intent-aligned rubric and credits any valid solution. Across seven frontier models, the strongest passes every checklist item on only 66.2% of tasks, and Design and Execution rank among the hardest stages for every model. Three models pass about 69% of tasks when only delivery contracts and general norms are checked, but just 28.4% to 40.0% once historical and material or workspace requirements are also checked. Evaluating delivery without these contextual requirements can therefore overestimate research agents and mask differences between them.

新论文 · v1arXiv:2611.13865

Online Network Design with Maximum Delay and Stochastic Arrivals

Tianhang Lu

原文摘要

We study online network services that combine pending requests into batches. Each batch pays for its spatial service and for the waiting time of its oldest request. The objective is the sum of these costs over all batches. Thus requests can share both a spatial purchase and one maximum-delay charge. We assume independent Poisson arrivals of known rates, while actual requests are revealed only upon arrival. For stateless services, suppose that each type has an attainable minimum cost when served alone, that every batch costs at least the stand-alone minimum of each type it contains, and that one finite-cost action can serve any batch, regardless of its request multiplicities. We give a deterministic algorithm with long-run ratio of expectations $e/(e-1)$ against an unrestricted offline optimum. This factor is optimal, even allowing randomization, already at one request location. The key comparison measures the union of time intervals associated with requests. The theorem applies directly to multi-root and pair aggregation on trees. We also study facility location, where each request pays its own connection distance and the bounded-action assumption can fail. A deterministic policy gives a ratio $3$ for arbitrary nonnegative opening costs. It immediately serves a finite initial set of requests, computes an assignment from their observed arrival gaps, and then keeps that assignment fixed. Its analysis combines Poisson insertion with a time-slice packing bound on the offline optimum. Every guarantee includes the full cost of completing finite inputs without an end-of-input signal.

cs.DS

新论文 · v1arXiv:2611.13859

Online Network Design with Additive Delays and Poisson Arrivals

Tianhang Lu, Runtian Ren, Shengcai Liu

原文摘要

We study online service systems in which requests arrive at locations in a network. A service may open a facility and connect customers, buy a subtree linking them to a root, or send a mobile server to visit them. Combining requests can save opening, network, or travel costs, but each request pays its own waiting time. The objective is the sum of these spatial costs and individual delays. We assume independent Poisson arrivals of known rates. The algorithms know how often requests occur, but observe them only upon arrival. We show that these rates suffice to make fixed spatial choices that compete with an offline optimum free to adapt to the realized input. We obtain long-run ratios of expectations below $27.249$ for facility location in finite metrics, $94.415$ for multi-root aggregation on trees, and $24.326$ for online service on trees. All three algorithms are deterministic; the multi-root algorithm also has polynomial-time preprocessing on rational inputs. The proofs connect rate-based spatial choices with lower bounds on unavoidable purchase, movement, and waiting costs. All main proofs are self-contained, and every guarantee includes the full cost of completing finite inputs.

cs.DS

新论文 · v1arXiv:2611.13757

Stochastic Aggregation with Prescribed Resources and Additive Delay

Tianhang Lu, Runtian Ren, Shengcai Liu

原文摘要

We study online aggregation with prescribed resources and additive delay. In this problem, we are given a finite set of resources with purchase costs and a collection of request types, each specifying the resources it requires. Requests arrive over time, and each request can be served immediately or postponed to a later time. A service may serve several pending requests together by purchasing every resource required by at least one of them, paying for each purchased resource only once. These purchases serve only the current batch; resources needed in a later service must be purchased again. A request served $t$ units of time after its arrival incurs delay cost $t$. The objective is to serve all requests while minimizing the sum of resource purchase costs and individual delay costs. We consider a stochastic version of this problem in which requests of each type follow an independent Poisson process of known rate. An online algorithm knows these rates but observes requests only upon arrival. We compare its expected cost with that of an offline optimum that knows the entire realized input and may choose every batch freely. A resource--scale linear program supplies both an offline lower certificate and synchronized service periods. This yields a deterministic online algorithm with polynomial-time preprocessing on rational input and a long-run ratio of expectations below $2.113$. The guarantee applies in particular to rooted multi-level aggregation and prescribed-pair aggregation on trees. Besides, for Poisson TCP acknowledgment, we further characterize the optimal long-run ratio of expectations at every load, showing that its worst-case value is approximately 1.409 and is attained by a deterministic online policy.

cs.DS

新论文 · v1arXiv:2611.13750

List Update with Per-Page Maximum Delay

Tianhang Lu, Runtian Ren, Shengcai Liu

原文摘要

We study list update with per-page maximum delay. Requests may wait, and one access clears all pending requests of a page at its current list rank, paying their maximum waiting penalty. Each service can change the permutation and therefore the cost of later accesses. We give a strictly $5$-competitive deterministic algorithm: serve a page when the oldest waiting time reaches two thirds of its current rank, then move it to the front. Our main tool is an asynchronous marked-rank coupling that compares different list orders and different service histories. A temporal transfer principle separates this permutation argument from two waiting-cost inequalities. It gives factor $5$ for page-specific superadditive penalties, common calendar rates, and additional hard deadlines; deadlines alone admit factor $3$. Two structural conditions on request-specific penalties give factor $3+2κ$. In particular, nonincreasing request severity retains factor $5$ even with unbounded weight ratios. For linear delay, offline optimization is NP-complete but fixed-parameter tractable in the number of pages. We characterize fixed-list batching, show an $Ω(m)$ gap between the best static order and dynamic reordering, and prove that the optimal deterministic factor for one page is exactly $2$. Counterexamples delineate the threshold rule's scope.

cs.DS

新论文 · v1arXiv:2611.12832

损失曲线遗漏了什么:FP4预训练减慢稀有事实的习得

What the Loss Curve Misses: FP4 Pretraining Slows the Acquisition of Rare Facts

Chang Lu, Mengjiao Zhang

AI 速读 基于摘要

作者报告,在验证损失几乎相同的情况下,NVFP4预训练模型习得稀有事实所需的曝光次数明显多于BF16。

cs.LG

查看原文摘要

Low-precision pretraining is usually judged by validation loss. We ask whether matching loss also means learning rare facts as efficiently. We train transformers from scratch with emulated 4-bit matrix multiplications on synthetic biographies with controlled exposure counts, on their own and mixed into web text. NVFP4 needs 1.32$\times$ and 1.52$\times$ as many exposures as BF16 to match its recall while validation loss differs by only 0.5% and 2%, and 1.26$\times$ and 1.30$\times$ when both models are evaluated with BF16 matrix multiplications. Under our primary readout, the exposure cost is about 1.2$\times$ at matched loss. Zero-mean gradient noise reproduces the cost of quantizing only the backward pass, which barely moves the loss. In a controlled experiment at twice the NVFP4 error, noise fed only to Adam's second-moment estimate acts like a lower learning rate and its cost fades, whereas noise in the update lowers its signal-to-noise ratio and carries about two thirds of the cost through the end of training. Raising the learning rate to offset the smaller average step does not recover the cost, either for this noise or for NVFP4. Gradient statistics of BF16 checkpoints rank the costs of formats, partial quantizations and noise controls without fitting (Spearman 0.96). We suggest measuring rare-fact acquisition alongside loss when evaluating low-precision recipes.

版本更新 · v2arXiv:2610.26507

Learn the Task, Not the Schedule: Reducing Spurious Cross-Arm Coupling in Bimanual Imitation Learning

Shuyuan Hu, Deng-Ke Wu, Yifan Jin, Qianhang Luo, Mingchen Ju, Kai Ye, Bo Zhao

原文摘要

A bimanual demonstration specifies both what the robot should do and one particular schedule for when each arm should act. Standard imitation learning does not distinguish task-required cross-arm constraints from incidental scheduling choices, and may therefore exploit opposite-arm progress as a shortcut cue, creating spurious cross-arm coupling even where relative timing is irrelevant for the current task phase. We introduce Constraint-preserving Temporal Recombination (CTR), a training-time data intervention that preserves recorded arm-local sensorimotor content while retiming the corresponding segments to realize admissible schedules that preserve task-required cross-arm constraints. We use CTR to train Diffusion Policy, Action Chunking with Transformers, and $π_{0.5}$, and evaluate them on four simulated and four real-robot tasks. Across six real-robot bimanual progress-state initializations, CTR raises worst-state success from 56% for the strongest non-CTR schedule mixture to 76%. With $π_{0.5}$ on an Independent task, a 4 s pause transfers 2.56 s of delay to the opposite arm under the Concurrent schedule distribution, but only 0.25 s with CTR. On real robots, CTR violates task constraints in only 0--4% of trials, versus 36--98% for narrow-schedule baselines. Code will be released at https://github.com/MINT-SJTU/CTR.

cs.RO · cs.AI · cs.LG

新论文 · v1arXiv:2611.12405

PTC-IoU: CPU-Friendly Detection Quality Estimation for Edge Video Analytics

Wei Geng, Hyerin Kim, Chang Lu, Zhenghao Lu, Joerg Ott

原文摘要

An edge video analytics system must estimate the quality of local detections since there are no ground-truth labels available during runtime. Detectors usually provide confidence scores for their results, but these scores can not reveal every missed object or misplaced bounding box. So a quality estimator is necessary to provide a decision signal indicating which frame has potential detection failures that need to be escalated to a more powerful model in the cloud. Neural network based estimators add workload to the GPU already running detection, while CPU capacity may remain available. We present Predictive Temporal Consistency IoU (PTC-IoU), which moves this workload to CPU by compacting detection boxes histories to support deterministic result scoring. Comparing to neural counterparts, experiments on representative edge hardware, Nvidia Jetson Orin Nano, show that our method reduced per-frame estimator latency from 10.3 ms for neural counterparts to 2.2 ms (4.7x faster), increased pipelined throughput by 13.6%, and cuts estimator energy consumption by up to 95%, while maintaining competitive final detection quality verified on multiple datasets.

cs.CV · cs.PF

新论文 · v1arXiv:2611.10457

Represented but Not Used: Value-Policy Dissociation in LLM Agents

Yuzhang Luo, Mohamed Aghzal, Liangming Pan, Ziyu Yao

原文摘要

Large language model (LLM) agents need to choose actions whose consequences unfold over multiple future steps, yet poor decisions do not reveal whether the model failed to represent those consequences or failed to use them. Mechanistic interpretability has studied internal representations of environment state and plans, but sequential decision making also requires evaluating how candidate actions change future outcomes. Reinforcement learning formalizes this as action value. We study this computation in Qwen2.5-7B-Instruct, Qwen2.5-32B-Instruct, and GPT-OSS-20B across a partially observed Grid environment and WebShop, a simulated web shopping environment. We find that the relative values of candidate actions are linearly decodable from intermediate hidden states, including at states where the policy prefers a worse action. We call this gap between value representation and policy choice value-policy dissociation. To investigate how represented value relates to action selection, we compare direct interventions along the value readout direction with Natural Value Transport (NVT), a transformation learned from hidden state changes that naturally result from differences in action value. While readout-direction interventions have limited and inconsistent effects, NVT more consistently influences action preferences, particularly at later layers. We also test a closed-loop controller that uses the value readout as a sensor and NVT as an actuator, observing improvements in task outcomes. These results suggest that LLM agents can represent useful prospective value without reliably translating it into policy choice.

cs.AI

版本更新 · v2arXiv:2607.03236

TACG: Trajectory-Aware Commit Gating for Diffusion Language Model Decoding

Chengcheng Wang, Tingzhang Luo, Wenhao Li, Jianyuan Guo, Chang Xu

原文摘要

Diffusion language models predict many masked positions at each denoising step, but their efficiency and output quality depend on which predictions enter the evolving context. Confidence is informative but incomplete for this decision: at matched confidence, proposals with different identity and support histories exhibit different one-step revision rates. We propose Trajectory-Aware Commit Gating (TACG), a training-free decoder that separates the current token proposal from the decision to commit it. TACG scores temporal support for the current proposal by contrasting its logits with an exponential moving average of earlier logits, requires the proposal identity to persist through a History Gate, and promotes a capped number of additional positions per step. The gate decides which positions to commit and when; every written token remains the contemporaneous base proposal. On complete four-task LLaDA evaluations under the stated single-seed protocol, measured task-specific operating points reach 51.02\% mean accuracy with 67.82 mean model evaluations, compared with 48.00\% and 76.59 for Native Confidence. Under a matched per-event commit-count replay, history-guided selection exceeds confidence ranking by 8.3 accuracy points at identical per-sample NFE. History-reference interventions alter commitment timing while preserving the base-proposal write rule. Experiments on Dream and LLaDA2-Mini characterize quality--cost behavior across backbones and generation lengths. The code is publicly available at https://github.com/Clarence-CV/TACG-DLLM.

cs.CL

新论文 · v1arXiv:2611.06627

Interactive World Simulation with Cosmos for Physical AI

NVIDIA, :, Anis Ahmad, Jon Allen, Malik Aqeel Anwar, Jesse Archer, Maciej Bala, Piotr Barejko 等 90 位作者

原文摘要

We present a shared framework that adapts Cosmos world foundation models for real-time interactive world simulation in Physical AI. By streaming visual observations in response to actions and scene controls, the framework enables users and embodied agents to interact with evolving virtual environments. It supports both predicting the visual consequences of actions and generating observations from physical state maintained by an external simulator. We demonstrate this approach across camera navigation and robot manipulation, controllable video transfer, and multi-view autonomous vehicle (AV) simulation. These examples provide a shared foundation for developing interactive virtual environments for Physical AI training and evaluation. Our code and model checkpoints are available at https://github.com/nvidia/cosmos and https://huggingface.co/collections/nvidia/cosmos3-examples. The project website is available at https://research.nvidia.com/labs/cosmos-lab/cosmos-simulation.

cs.AI · cs.CV · cs.LG · cs.RO

新论文 · v1arXiv:2611.05006

When Category Means Move: A Paired-Support Audit of Recommendation Logs with Random Insertions

Yu Zhang, Zhihan Wang, Yuhang Lu, Shuai Li

原文摘要

Content-category dashboards summarize production recommendation logs, although the serving policy selects the users, items, and contexts behind each mean. Using KuaiRand random-insertion impressions, we audit these summaries on the same 12,472 users and 5,486 videos in both logs. Standard means have absolute/centered mean absolute error (MAE) 0.1935/0.0491 and rank correlation 0.7833. User demeaning plus a random grand-mean anchor reduces absolute MAE by 80.0%, but centered MAE by only 21.0%, without reliably improving rank. A complete-support ledger allocates weighted squared error mainly to within-video residual (72.41%) and video mix (23.90%), with 3.65% from random support. In a frozen later test, early category calibration reaches MAE 0.0251 versus 0.0531 for scalar calibration; their direct difference is -0.0280 [-0.0296, -0.0172]. Decision evidence is cutoff-specific: the nominal top-10 regret change for row-weighted demeaning versus naive is -0.0062 [-0.0084, -0.0003], while top-5 and top-20 intervals include zero. The six-contrast family-wise top-10 interval [-0.0136, 0.0013] includes zero. An outcome-free disjoint randomized-reference split also leaves all six nominal contrasts covering zero. A known-truth stress test shows incomplete improvement under a mean-additive positive control and failure under item mix and weak overlap. All decompositions are descriptive and each result is conditional on its overlap cohort. KuaiRand intervals quantify user-cluster variation conditional on the realized catalog and calendar; we claim neither intrinsic category effects nor a universally valid correction.

cs.IR

新论文 · v1arXiv:2611.04997

Auditing Multi-Proxy Explanations of Web User Preferences: Identification Boundaries and Abstention

Yu Zhang, Zhihan Wang, Yuhang Lu, Shuai Li

原文摘要

Web systems infer user preferences from behavior, ratings, text, and LLM-generated profiles, yet explaining a proxy may reveal its measurement process rather than the preference of interest. We develop an audit framework distinguishing point identification, partial identification, and abstention relative to an explicit reference distribution. Under feature-dependent, arbitrarily signed proxy bias, a strict majority of uncontaminated proxy attributions lets the coordinatewise median recover latent-attribution support and direction and, with externally justified positive scales, magnitude. We prove this threshold sharp using response-valid indistinguishable worlds, characterize sharp minimax error and exact compatible-target sets under approximate cleanliness and scale uncertainty, and propagate clean-proxy attribution error. A frozen known-truth benchmark spans two response families, four proxy counts, all contamination counts, three bias strengths, and eleven methods, with a focal comparison confirmed on twenty unseen seeds. Under correct budgets, a direct attribution-tensor audit achieved 100% coverage, 95.40%-97.05% active-driver sign decisions, and no false signs. In a temporally separated Amazon study, median aggregation reduced comparator-relative conditional sign error for all four targets; six of eight prespecified target-metric contrasts had Bonferroni intervals excluding zero. A controlled real-covariate perturbation favored the median on the prespecified focal metric for all four targets. Compatible sets showed that the current proxy configuration rarely licensed strict-sign conclusions. Weak comparator calibration limits the Amazon evidence to fitted-surface alignment rather than latent-preference recovery.

cs.LG · cs.IR

版本更新 · v2arXiv:2609.35290

EvoIn: Bridging Evolution and Internalization for Agent Fine-Tuning

Shihan Dou, Shaofan Liu, Zhonghang Lu, Jiahang Lin, Shichun Liu, Binghai Wang, Jiajie Jin, Guanting Dong 等 11 位作者

原文摘要

Recent work has explored improving agents by jointly evolving their harnesses and models, but often takes a ''potpourri'' approach that bundles together new tools, new decision-making procedures, and model adaptation to the evolved harness under a single notion of agent improvement. In this paper, we instead investigate how agents can improve their decision-making procedures. In particular, we propose EvoIn, an agent fine-tuning framework that bridges evolution and internalization. EvoIn first analyzes agent execution traces to evolve and validate new decision-making procedures by temporarily instantiating them in the harness. The validated procedures guide the agent to generate improved reasoning traces. These traces are then rewritten into self-contained reasoning traces, removing explicit references to harness instructions while expressing the induced decision logic as the model's own reasoning. Finally, EvoIn fine-tunes the model on the rewritten traces, internalizing these procedures so that the improved decision-making persists without the evolved harness at inference time. We evaluate EvoIn on diverse benchmarks and find that it consistently enables agents to learn stronger decision-making procedures, raising the pass rate by 10.9 points in-domain and by 9.2 points out-of-domain. Results further show that the internalized decision procedures generalize to unseen tasks. Case studies show that agents can learn to decide how to solve a task before solving it, for example by checking a document's length to choose between reading it in full and searching it. EvoIn is also broadly applicable, showing consistent improvements on another model family.

cs.AI · cs.CL

版本更新 · v3arXiv:2608.18577

Online Service with Per-Batch Maximum Delays

Tianhang Lu, Runtian Ren, Shengcai Liu, Ke Tang

原文摘要

We study online service with per-batch maximum delays on a finite metric space. Requests arrive over time, and a server serves each batch by following a walk visiting its requests' locations. Movement is instantaneous. The objective is to minimize the total distance traveled by the server plus the sum, over all batches, of the maximum waiting time in each batch. We give deterministic polynomial-time algorithms with competitive ratios $10$ on finite lines, $12$ on weighted trees, and $20$ on arbitrary finite metrics. Unlike additive delay, maximum delay allows younger requests to join a batch without increasing its delay charge. Our algorithms coordinate service using a common waiting clock and a threshold determined by the metric. The analysis uses certificates for groups of requests to compare this policy with any offline schedule, even one that serves different batches and follows different walks. This comparison shows that coordinating service in this way loses only a constant factor. The bounds hold whether the server must serve requests it visits or may leave them unserved; the two rules also have equal offline optimal costs. Offline optimization is polynomial on lines and explicitly represented weighted trees, and NP hard on general metrics. When arrivals are announced but locations are hidden until visited, we give an $84$-competitive algorithm on a known finite line under either rule.

cs.DS

版本更新 · v4arXiv:2608.06796

Online Multi-Level Aggregation with Maximum Delay

Tianhang Lu, Runtian Ren, Shengcai Liu, Ke Tang

原文摘要

We study Multi-Level Aggregation with Maximum Delay (MLA-MD) on finite weighted rooted trees. Each service pays for a rooted subtree and for the maximum waiting time among the requests it clears. For the offline problem, we give an exact polynomial-time algorithm that computes both the optimum and an optimal service schedule. For the online problem, we give a deterministic $2$-competitive algorithm and, against an oblivious adversary, a randomized $e/(e-1)$-competitive algorithm. Both competitive ratios are optimal, even on rooted paths. We also extend the offline result and both online guarantees to service systems with submodular costs.

cs.DS · cs.DM · math.OC

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.2058ACL Anthology · ACL 2026

AgentGym2: Benchmarking Large Language Model Agents in De-Idealized Real-World Environments

Zhiheng Xi, Dingwen Yang, Jiaqi Liu, Jixuan Huang Fudan University, Honglin Guo Fudan University, Baodai Huang, Tinggang Chen, Qi Zhang 等 25 位作者

原文摘要

Language agents, i.e., LLM agents, progress rapidly and are increasingly deployed in production environments. This trend underscores the urgent need for rigorous and realistic evaluations. However, most existing benchmarks evaluate agents in simplified, idealized settings. They typically rely on pre-packaged tool interfaces, overlook critical steps, and assume inputs are clean and fully specified. Consequently, they understate the difficulty of real deployments, where uncertainty and noise are ubiquitous and agents must proactively explore the environment to uncover new tools. To bridge this gap, we present AgentGym2, a new evaluation framework with task instances grounded in real-world end-to-end working demands. Beyond reasoning and planning, it measures agents’ ability to execute end-to-end procedures, discover tools via exploration, compose tools for unseen tasks, and remain robust to noisy and underspecified information. Experiments on 15 proprietary and open-source models show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2, revealing a substantial gap between the capability of current agents and the demands of real-world applications.

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.681ACL Anthology · ACL 2026

Min-k Sampling: Decoupling Truncation from Temperature Scaling via Relative Logit Dynamics

Yuanhao Ding, Meimingwei Li, Esteban Garces Arias Ludwig-Maximilians-Universität München, Matthias Aßenmacher Ludwig-Maximilians-Universität München, Christian Heumann Ludwig-Maximilians-Universität München, Chongsheng Zhang Ludwig-Maximilians-Universität München and Henan Univeristy

原文摘要

The quality of text generated by large language models depends critically on the decoding sampling strategy. While mainstream methods such as Top-k, Top-p, and Min-p achieve a balance between diversity and accuracy through probability-space truncation, they share an inherent limitation: extreme sensitivity to the temperature parameter. Recent logit-space approaches like Top-n\sigma achieve temperature invariance but rely on global statistics that are susceptible to long-tail noise, failing to capture fine-grained confidence structures among top candidates. We propose Min-k Sampling, a novel dynamic truncation strategy that analyzes the local shape of the sorted logit distribution to identify “semantic cliffs”: sharp transitions from high-confidence core tokens to uncertain long-tail tokens. By computing a position-weighted relative decay rate, Min-k dynamically determines truncation boundaries at each generation step. We formally prove that Min-k achieves strict temperature invariance and empirically demonstrate its low sensitivity to hyperparameter choices. Experiments on multiple reasoning benchmarks, creative writing tasks, and human evaluation show that Min-k consistently improves text quality, maintaining robust performance even under extreme temperature settings where probability-based methods collapse. We make our code, models, and analysis tools publicly available.

cs.CL

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