跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+0新增论文
+0其中 AI 相关
+0已有论文修订
+92新增 AI 导读

arXiv 最近完成扫描:2026-11-23 01:07 · 全站最近入库:2026-11-21 12:19各信源进度 →

统计口径与数据新鲜度

统计 2026-11-22 03:23 — 2026-11-23 03:23(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:2026-11-23 01:07最近发现新增版本:2026-11-21 12:19库内论文最新更新时间:2026-11-20 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。

29,158 已收录论文24,181 AI 相关8,992 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 324 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-23 01:05(北京时间)
新论文 · v1arXiv:2611.26427

MemFolder:面向高效智能体记忆管理的结构引导记忆折叠

MemFolder: Structure-Guided Memory Folding for Efficient Agent Memory Management

Xuyang Liu, Chenwen Lu, Junxi Wang, Siyu Ren, Xinying Lin, Siteng Huang, Hongli Sun, Linfeng Zhang 等 9 位作者

AI 速读 基于摘要

MemFolder 通过结构引导的记忆折叠,依次减少记录数量和上下文长度,以降低长期智能体的存储、检索与生成成本。

cs.CL · cs.CV

查看原文摘要

Long-term agents accumulate large external memory stores as interactions unfold, increasing the costs of storage, retrieval, and response generation. Efficiency is constrained by two coupled factors: a growing record count increases retrieval overhead, while longer retrieved contexts increase prefill computation and decoding-time attention costs. Existing methods prune or consolidate records, but pruning can discard useful evidence, whereas consolidation may produce fewer yet longer retrieval units. We introduce MemFolder, a structure-guided memory folding framework that sequentially reduces record count and context length. The framework comprises two core components that share memory groups, using cross-record structure to guide both evidence consolidation and compact representation. RecordFolder grounds entity mentions, progressively selects representative yet complementary anchors, and combines semantic and entity affinity with temporal chaining to consolidate compatible records into fewer retrieval units. ContextFolder encodes the original records within these groups into compact text and uses verification feedback to revise encodings when key facts cannot be recovered. Experiments on the three text-based memory benchmarks LoCoMo, PersonaMem, and LongMemEval and the two multimodal benchmarks M3-Bench-web and M3-Bench-robot demonstrate the strong performance of our methods. Notably, on Mem0, RecordFolder retains 99.0% of the original average performance across the three text-based benchmarks with 50% fewer records, exceeding the strongest baseline by 1.9 percentage points in performance retention. Combined with ContextFolder, existing record compression methods further shorten their output records while retaining most of their performance.

新论文 · v1arXiv:2611.26358

Beam 模型技术报告

Beam Model Technical Report

Reflection Team, Adam Abdulhamid, Osama Abuelsorour, Yasmin Alameddine, Carlo Alfano, Duarte Alves, Brandon Amos, Selvaraj Anandaraj 等 230 位作者

AI 速读 基于摘要

Beam 是一个面向编码、推理与智能体工作负载的稀疏 MoE 模型,总参数 501.3B、激活参数 23.4B,作者报告其推理算力需求比同类模型低 3–4 倍。

cs.CL · cs.AI

查看原文摘要

Beam is a sparse Mixture-of-Experts (MoE) model with 501.3B total parameters and 23.4B active parameters, specifically built and optimized for coding, reasoning, and agentic workloads. Beam delivers frontier level reasoning, while requiring 3-4x less inference compute than many models of its class. Pretrained on 23.8T diverse, curated, high-quality tokens, Beam features a 1M-token context window achieved through a novel two-stage midtraining phase, built on verifiable, correctness-anchored data. Post-training leveraged a large-scale, fully asynchronous reinforcement learning (RL) campaign across 10.5K GB300 GPUs, generating over 100M rollouts over 4 weeks of training. We believe this is one of the largest scale RL runs conducted by any open lab to date. Across our evaluation suite, capabilities continued to improve as we increased RL compute, with no sign of a plateau. Supported by deep hardware-model co-design, FP8/NVFP4 quantization, and a 1.93B-parameter speculative decoding drafter (DFlash 2), Beam maximizes serving throughput with efficient, flexible deployment. Averaged across conversation turns and harm categories, Beam refuses or pivots on 88.9% of harmful requests while complying with 81.5% of harmless requests, achieving the highest refusal/pivot rate on harmful requests among open-weight models on the Pareto frontier.

新论文 · v1arXiv:2611.26275

dLoop:通过循环潜在精炼的循环扩散语言模型

dLoop: Looped Diffusion Language Models via Recurrent Latent Refinement

Handa Li, Pengxiang Li, Mingyu Cao, Xilu Wang, Li Shen, Da Li, Lu Yin

AI 速读 基于摘要

dLoop 将预训练扩散语言模型重构为 Prelude-Loop-Coda 架构,通过循环应用参数共享的中间块来增强潜在计算,从而在承诺前更充分地整合信息,提升生成准确率与效率。

cs.CL

查看原文摘要

Diffusion language models (DLMs) support parallel generation, yet they often require many denoising steps because only a fraction of token predictions are sufficiently reliable to be committed at each step. Existing acceleration methods primarily optimize confidence learning or commitment policies. However, these approaches operate on the outputs of a single denoising forward pass, in which information from the prompt, previously committed tokens, and uncommitted positions may not yet be fully integrated. Consequently, some predictions remain under-refined when the commitment decision is made, and modifying the commitment policy alone cannot improve their underlying representations. We introduce dLoop, which restructures a pretrained diffusion language model into a Prelude-Loop-Coda architecture and recurrently applies a parameter-shared intermediate block to the same partially masked state. Rather than advancing the denoising process, each additional loop provides further latent computation, allowing information distributed across uncommitted positions to continue interacting and become more fully integrated before commitment. Experiments on GSM8K, MATH, MBPP, and HumanEval show that dLoop achieves the highest or tied-highest accuracy on every benchmark. With tokens per forward pass (TPF) adjusted for loop computation, it outperforms Fast-dLLM in both accuracy and TPF on all four benchmarks and dParallel on three, and is more accurate than d3LLM on all four, while d3LLM attains higher TPF; no compared method surpasses dLoop in both. Mechanistic analyses show that extra loops mainly help on slow, unstable trajectories and that updates at uncommitted positions drive changes in token decisions and later generation. Pre-commitment latent computation thus makes predictions more reliable, offering a route to DLM acceleration distinct from forcing confidence through the training loss.

版本更新 · v3arXiv:2606.27786

SPARK: Adaptive Parametric Knowledge Participation for Faithful Retrieval-Augmented Generation

Ruochang Li, Pengcheng Huang, Zhenghao Liu, Yukun Yan, Huiyuan Xie, Yu Gu, Ge Yu, Maosong Sun

原文摘要

Retrieval-augmented generation (RAG) enhances the faithfulness of large language models (LLMs) by incorporating external knowledge into the generation process. However, conflicts between retrieved context and parametric knowledge can cause LLMs to rely on inconsistent internal knowledge, hindering their ability to faithfully leverage retrieved evidence. Existing approaches mitigate this issue by editing parametric knowledge to resolve such conflicts. However, locating and modifying knowledge encoded in model parameters is challenging and may inadvertently compromise the backbone LLM's capabilities. In this paper, we introduce SPARK, a novel framework that formulates parametric knowledge control as a latent participation policy learning problem, enabling LLMs to adaptively regulate the contribution of parametric knowledge during generation. Specifically, SPARK introduces a lightweight participation policy that dynamically modulates the contribution of FFN-mediated parametric knowledge while keeping the backbone LLM frozen. The participation policy is optimized with a faithfulness-oriented reward, enabling fine-grained control over parametric knowledge participation during generation. Extensive experiments on six datasets validate the effectiveness of SPARK, which introduces and optimizes only 0.01\% additional trainable parameters while keeping the backbone LLM fully frozen, thereby preserving its general capabilities and achieving more faithful generation than competitive baselines. All datasets and code are available at https://github.com/OpenBMB/SPARK.

cs.CL · cs.AI

新论文 · v1arXiv:2611.26000

Xing-Audio-NVVASR: Scaling and Refining Supervision for Joint Recognition of Speech and Non-Verbal Vocalizations

Pengyu Xu, Gang Liu, Genliang Zhao, Hongjie Chen, Jian Kang, Jie Lian, Jie Li, Yongxiang Li

原文摘要

Joint recognition of speech and non-verbal vocalizations (NVVs) is hindered by scarce event-rich training data and inaccurate or missing event labels. We present Xing-Audio-NVVASR, an end-to-end recognizer trained through progressive data mining and iterative supervision refinement. Guided by unified acoustic criteria, two heterogeneous models cross-audit annotations, flagging disagreements for teacher reannotation. To uncover shared omissions that model agreement may conceal, complementary breath detectors screen the entire corpus for human review. Human-confirmed labels accumulate across rounds and are preserved during subsequent refinement. Over five rounds, we refine annotations for 170,000 utterances from public corpora and incorporate 630,000 mined utterances. The resulting corpus of approximately 800,000 utterances is used to train a single 1.7B-parameter model that transcribes speech with inline tags for six NVV classes. Across four human-annotated benchmarks, Xing-Audio-NVVASR ranks first among nine models in both composite score and six-class Macro-F1. It achieves a mean composite score of 64.46, surpassing the strongest comparison model, StepAudio 3 Chat Preview, by 6.03 points, and exceeds the annotation teacher's breathing F1 by 14.40 to 31.57 percentage points. We plan to release the model weights, inference code, four evaluation benchmarks, annotation guidelines, and evaluation code.

cs.CL

版本更新 · v5arXiv:2507.18061

TELEVAL: A Benchmark Designed for Spoken Language Models in Chinese Interactive Scenarios

Zehan Li, Hongjie Chen, Qing Wang, Yuxin Zhang, Jing Zhou, Hang Lv, Mengjie Du, Yaodong Song 等 12 位作者

原文摘要

Spoken Language Models (SLMs) are expected to support natural spoken interaction beyond task completion. Existing benchmarks assess a range of semantic tasks and conversational capabilities. However, high task scores do not necessarily imply reliable and appropriate spoken interaction. To address this gap, we introduce TELEVAL, a large-scale benchmark focused on Chinese spoken interaction. It evaluates responses to spoken inputs without additional task-specific prompts or predefined answer options. TELEVAL covers two complementary aspects: (1) Reliable Content Fulfillment, which measures semantic accuracy under diverse acoustic and linguistic conditions, and (2) Interactional Appropriateness, which assesses conversational naturalness and appropriate adaptation to auditory cues. Experiments reveal limitations in both aspects. Strong results on factual question-answering datasets do not consistently extend to locally relevant knowledge or challenging speech conditions, such as Chinese cultural questions and dialectal speech. Models also struggle to adapt their responses to auditory cues, even in cases where they explicitly recognize them. We further document a recurring failure pattern, termed the "Caption Trap", in which some responses describe a perceived auditory cue instead of addressing the user's conversational needs. Overall, the evaluated SLMs struggle to deliver consistently accurate and conversationally appropriate responses. TELEVAL provides a diagnostic framework for evaluating both content reliability and conversational adaptation.

cs.CL · cs.AI · cs.SD · eess.AS

新论文 · v1arXiv:2611.25582

AutoRAGBench: A Benchmark for LLM Agents Automating RAG System Development

Hang Zhang, Yuzhi Pan, Ziyao Zhong, Chaokun Wang, Shuo Cao, Zeyu Huang, Yue Xue, Fang Niu 等 15 位作者

原文摘要

Retrieval-augmented generation (RAG) is a well-established paradigm for augmenting large language models (LLMs) with information retrieved from external corpora. However, developing effective RAG systems requires substantial engineering effort and iterative experimentation, as effective design choices vary across tasks and domains. With rapid advances in LLM agent capabilities, we ask how well these agents can autonomously build RAG systems and how effectively they can improve them through iteration. In this paper, we propose AutoRAGBench, the first benchmark to evaluate how well LLM agents develop RAG systems from scratch in an open code space. It comprises nine RAG system development tasks. For each task, agents build RAG systems from scratch without starter code or prescribed strategies, under a six-hour budget on a single H800 GPU. They analyze task data, implement RAG pipelines, and iteratively refine their designs based on experimental feedback. We evaluate nine agent configurations and analyze both the effectiveness of the resulting RAG systems and their development trajectories. GPT-6.1 Sol with Codex CLI achieves the highest average score of 56.1. The composite Human-SOTA baseline averages 44.4. These results show that LLM agents can automate effective RAG system development. When starting from a human-designed system on each of seven tasks, GPT-6.1 Sol with Codex CLI improves the score by 16.6 points on average.

cs.CL · cs.AI · cs.IR

版本更新 · v4arXiv:2609.12541

Agent as Policy for Robotic Manipulation

Mengzhao Jia, Yang Lin, Xixin Zhang, Zhihan Zhang, Meng Jiang

原文摘要

We demonstrate that a general-purpose coding agent can directly drive a physical robot throughout task execution without any task-specific or environment-specific training. We introduce Agent as Policy (AGP), which places task planning and execution under the agent's control. Given a task and a robot interface, the agent interprets visual evidence, writes executable programs, issues motion commands, and revises its actions in response to physical outcomes. This brings the agent's reasoning and programming capabilities into continuous interaction with the physical world. We study AGP across multiple real-world manipulation tasks spanning precision manipulation, dynamic motions, and deformable objects. These include assembly from human videos, block construction from goal images, dice flipping, targeted throwing, and bimanual towel folding. AGP achieves success rates of at least 80% in seven of eight task configurations and significantly outperforms previous agentic robot systems. We further study efficiency through task experience accumulation and find that reusing saved procedures and programs shortens execution time across repeated trials. These findings support a path for general-purpose agents to act as robot policies, extending their autonomy to physical manipulation through runtime reasoning, programming, and interaction.

cs.CL

新论文 · v1arXiv:2611.24845

Vibe-Trading:迈向智能体原生交易

Vibe-Trading: Towards Agent-Native Trading

Haozhe Wu, Zhonghang Li, Chao Huang

AI 速读 基于摘要

提出一个面向金融交易的智能体原生框架与一体化工作区,通过保留结果来源、时间范围与执行权限,使研究到执行的过程可追溯、可复现并受显式授权约束。

cs.AI · cs.CL · cs.MA · q-fin.CP · q-fin.TR

查看原文摘要

Existing AI agents can retrieve information, call tools, run experiments, and produce multi-step outputs. Yet most are built around relatively static workflows and struggle to preserve reliable state as the environment changes. Financial trading makes this limitation especially consequential: prices, news, market conditions, portfolio positions, and available capital evolve continuously. A result is useful only when it remains tied to the correct time, data snapshot, and execution context. We present Vibe-Trading, a comprehensive agent-native trading framework and all-in-one workspace for building, evaluating, and evolving agents for financial trading. It unifies market research, financial data, strategy generation, backtesting, portfolio analysis, paper trading, multi-agent collaboration, and optional broker execution. Its central contribution is an architecture that keeps results, evidence, research history, and trading authority connected throughout the research-to-action process. Each result retains its source, temporal scope, configuration, and producing run. Shared evidence grounding enables verification, versioned research records support reproducible continuation, and an explicit authority gate separates strategy research from live execution. Together, these designs provide a practical foundation for trading agents whose results can be traced, reproduced, and acted on only under explicit authority.

新论文 · v1arXiv:2611.24651

Piccolo: Co-Evolving Decision Models and Data through Recursive Self-Improvement

Zhenqing Ling, Qirui Jiao, Daoyuan Chen, Cheng Chen, Xinji Mai, Haowen Gao, Zhuo Zhang, Xuchen Pan 等 17 位作者

原文摘要

We present Piccolo, a family of multimodal decision models (0.8B to 27B) that coevolves model capabilities and training data via Agentic Data Engineering and recursive self-improvement (RSI). The models predict probabilities over candidate answers for general-purpose text and visual judgments, agent harnesses, data workflows, and interactive games. Piccolo-27B tops the Decision Index Vision Full leaderboard with a score of 75.59, while Piccolo-9B and Piccolo-4B achieve 68.35 and 67.46, respectively. All three rank first on the Private benchmark in their respective model-size categories. Research agents analyze model behavior, construct training data, conduct experiments, and use the results to refine subsequent proposals. Semantic slices and a shared research memory connect these actions across the model family's development. With training recipes comprising hundreds of thousands of examples, data selection and feedback-driven refinement are central to the self-improvement process. Our large-scale RSI exploration yields three counter-intuitive insights for decision-model development: (1) Backbone-recipe co-adaptation outweighs raw model recency: earlier backbones can surpass newer generations on visual tasks under identical recipes; (2) Targeted data curation enables 4B-9B models to rival 27B baselines on specialized workflows; and (3) Continuous loss reduction can decouple from and even degrade discrete decision accuracy. Together with the released weights and tools from the Data-Juicer and AgentScope ecosystems, Piccolo provides an actionable blueprint for building frontier decision models and scaling data-driven RSI.

cs.CL · cs.AI · cs.CV

版本更新 · v2arXiv:2609.24657

Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models

Xiaoqiang Wang, Mengyang Xiong, Jun Dai, Bang Liu

原文摘要

Improving pretrained language models with limited adaptation data calls for computational modules that are both effective and scalable. Parameterized quantum circuits offer expressive transformations for language modeling, yet existing quantum-enhanced language models apply the same predefined circuit design to every token in an input sequence, regardless of its content. This limitation is particularly acute in masked-diffusion language models, where token representations evolve across iterative denoising steps while the circuit design remains fixed. We introduce HyperQ, a quantum-augmented masked-diffusion language model that dynamically synthesizes token-specific quantum circuits. Built upon an IQP-inspired architecture, HyperQ employs a lightweight circuit hypernetwork to generate rotation angles and measurement bases conditioned on each token's hidden state. The resulting quantum expectation values are projected back into the frozen language-model backbone through a residual connection. To support wider quantum registers, HyperQ bounds circuit interaction degree, maintaining per-qubit gradient scale and readout cost independent of register width. Across six zero-shot benchmarks, scaling HyperQ from 16 to 64 qubits improves average accuracy from 47.65 to 54.30, surpassing the frozen backbone's 49.59 by 4.71 points while using only one tenth of the fine-tuning data used by classical baselines. We further execute the generated circuits on an IBM 156-qubit superconducting quantum processor, demonstrating their feasibility on physical quantum hardware.

quant-ph · cs.CL

版本更新 · v3arXiv:2610.16352

LoongSage: A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes

LoongSage Team, Bowei Song, Caiwei Zhao, Chuwen Liu, Dechuang Gu, Dong Fang, Feiyu Xie, Huaiyu Li 等 48 位作者

原文摘要

Reinforcement learning (RL) has emerged as a critical phase in the post-training pipeline of large language models (LLMs). As the frontier of LLM capabilities shifts from static reasoning to agentic tasks--characterized by external tool invocation, multi-turn interactions, and long-horizon decision-making--existing mainstream RL frameworks have grown increasingly unwieldy, raising the overhead of adoption, usage, and debugging. To address these challenges, we present LoongSage, an agent-native, fully asynchronous RL framework with a lightweight core of roughly 14K lines of code. Built upon Megatron for distributed training and SGLang for high-throughput inference, LoongSage centralizes orchestration within a single controller process and encapsulates all GPU-resident components as Ray actors. It introduces AgentFlow, a unified execution paradigm that orchestrates both white-box and black-box agents. For asynchronous execution, it integrates sliding-window strategies that bound trajectory staleness, and employs TransferMesh, a topology-aware synchronization mechanism that routes weights after each step over the fastest transport available to each receiver. Beyond agentic RL, LoongSage incorporates a memory-efficient Teacher Manager that decouples the scaling of teacher models from GPU memory constraints, natively supporting Multi-Teacher On-Policy Distillation. Extensive evaluations across four complete training runs, spanning complex reinforcement learning and multi-teacher distillation, demonstrate that task rewards improve steadily while the train-inference gap is kept under continuous control, and the fully asynchronous pipeline removes synchronous idle bubbles, sustaining hardware utilization.

cs.LG · cs.CL

版本更新 · v2arXiv:2611.00515

AStockArena: Beyond Returns, Can We Entrust Real Capital to the Decisions of Financial Agents?

Zeyu Liu, Sing Kwong Lai, Xuanzhi Feng, HaoYang Shang, Yang Li, Zihao Fang, Xuan Liu, Song Guo

原文摘要

Large language models (LLMs) are increasingly used in dynamic financial trading, yet real-time trading benchmarks mainly evaluate agents by cumulative returns and risk-adjusted performance. Such outcome metrics cannot show whether trades are faithfully reported, supported by available evidence, and consistent with account states. We introduce AStockArena, a real-time decision-auditing benchmark for continuous trading in China's A-share market. Under shared market conditions and execution constraints, we evaluate 13 open- and closed-source models over 3,198 sequential decisions. AStockArena reconstructsthe actions actually posted to accounts by aligning reports, trade requests, execution responses, and account states, and audits them with eight checks spanning execution traceability (O), evidence sufficiency (RK), and financial meta-awareness (B), combining record verification with consensus expert annotation. Of 1,070 trading points, 72.99% exhibit at least one audit gap; model-level rates range from52.2% to 91.2%, with evidence gaps most prevalent. Historical trade-filtering replay further shows thatportfolio outcomes are sensitive to these flagged trades. Decision-process auditing thus offers a verifiable complement to return-based evaluation of financial agents.

cs.CL · cs.AI

新论文 · v1arXiv:2611.23570

Safe to Omit, Worth Omitting? Keeping the Answer and Cutting Latency When an LLM Agent Skips Another's Message

Xinyang Li, Chih-Hsuan Yang, Xiaolong Ma, Anjir Ahmed Chowdhury, Vikram Vasudevan, Rajeev Thakur, Feng Yan

原文摘要

In multi-agent systems, a receiver agent often waits for a helper agent's message (text or its latent KV cache), which can improve performance but adds latency. We study when a receiver agent can skip the message safely (the final answer rarely changes) and beneficially (the time saved exceeds the cost of deciding). Our mechanism reads the receiver agent's confidence without the message and uses a finite-sample test to certify a threshold at which, with high probability, at most 5% of skipped answers change; if none passes, the system keeps communicating. We test it on OpenBookQA, ARC-Challenge, and MMLU-Pro with three Qwen model pairs, on cross-family agents, and on SQuAD and GSM8K. Four insights emerge. First, oracle headroom is weak evidence that messages help: control inputs without answer-specific content, matched to change answers as often as real messages, reproduce a median 76% of the oracle's gain. Second, whether a message can be skipped depends on how often and how predictably it alters the answer, not on its accuracy; every certified KV-cache path is less accurate than the receiver agent alone. Third, where text messages help, skipping keeps most of their gain: on unseen questions in the four such settings, the policies skip 59 to 80% and keep an estimated 76 to 95% of the gain. In two prospective tests with all rules fixed in advance, our mechanism skips a three-agent system's verifier on 39.9% of sealed questions, keeping 12.55 of its 12.80-point gain, and a 32B helper on 68.4% of held-out questions, keeping its whole gain. Fourth, safety does not determine efficiency: at batch size one, skipping text saves hundreds of milliseconds per query and skipping KV cache far less. Certification extends to SQuAD but not to GSM8K. In eleven tests on unseen questions, every certified policy we could test kept its change rate's 95% upper bound below 5% under our answer parsers.

cs.LG · cs.CL

新论文 · v1arXiv:2611.23388

QDA: Closed-Loop LLM Quantization Design with Agents

Xingyang Li, Tianlang Zhao, Song Han, Ligeng Zhu

原文摘要

Post-training quantization is the standard way to serve large language models at low cost. However, the best quantization recipe differs across models and precisions. As a result, experts must redesign it for every new model, which is time-consuming and often sub-optimal. Fortunately, an LLM agent can take over this design by refining the recipe over rounds of building and evaluating candidates, so that more rounds lead to better recipes. However, evaluating one quantized reasoning model takes tens of GPU-hours, and combining methods from different codebases means porting each to the new architecture by hand. We present QDA, a system in which an agent designs the recipe for each model and precision. To make evaluation fast, QDA scores each candidate in minutes by its output KL divergence from the original model on reasoning traces that the agent never sees, and checks promising ones with long generations to catch failures such as repetition loops. To make building fast, QDA provides a library of public PTQ methods that are compatible with one another and with new model architectures. Across four reasoning models, QDA improves accuracy over the strongest public recipe by 0.9, 1.4 and 2.2 percentage points on average at W4A16, W4A8 and W3A16, using 64 GPU-hours per search. On the 320B-parameter MoE model GLM-5.3-Flash at NVFP4 W4A4, it recovers 98.3% of the FP8 model's accuracy, against 96.9% for the best vendor checkpoint. Every QDA checkpoint also runs on existing serving kernels with no inference overhead. We will release the library, the evaluator and the harness.

cs.LG · cs.AI · cs.CL

版本更新 · v2arXiv:2607.24651

Evidence Attribution in Visual Document Understanding without Coordinates or Region Labels

Zhuchenyang Liu, Yao Zhang, Yu Xiao

原文摘要

Reliable visual document understanding requires a model to attribute each answer to the evidence regions that support it. Recent benchmarks and systems express this step through a coordinate interface: the model outputs the coordinates of bounding boxes that mark the evidence regions in the document. Under this interface, vision-language models often fail to identify the right regions even when the answer is correct, a failure known as Attribution Hallucination. We present a study that investigates whether this failure is partially limited by what the model can express through coordinates. On a verified bilingual CiteVQA subset, we compare the coordinate interface with a language interface in which the model outputs only text, quoting its evidence verbatim, and a multimodal retriever returns the location of each quote as a page region proposed by a layout parser (tables and figures are quoted through their captions or notes); the comparison is repeated over six open vision-language models. Compared with the coordinate interface, evidence recall rises from at most 8 points to between 26 and 47 and the hallucination rate roughly halves, with little change in answer quality. Building on this comparison, we use the same quote-and-retrieve pipeline as a training scaffold: because region-level evidence labels are expensive to collect for long documents, we introduce a GRPO recipe whose reward is a judge's reading of the gold answer and crops of the retrieved regions, training the model to quote better evidence without any region labels and raising an 8B backbone's strict attributed accuracy from 22.4 to 33.8. These findings indicate a practical path to improve attribution without a coordinate interface and without costly region-level supervision.

cs.CV · cs.CL · cs.IR

新论文 · v1arXiv:2611.23357

A Survey on RLOps for Large Language Models

Lingzhe Zhang, Tong Jia, Chiming Duan, Minghua He, Leyi Pan, Kening Zheng, Zhaoyang Liu, Bolin Ding 等 10 位作者

原文摘要

Reinforcement learning (RL) has become increasingly important for improving the reasoning, instruction-following, and task-solving capabilities of large language models (LLMs). However, scaling RL training exposes failures in distributed execution, feedback computation, and learning dynamics that can waste computational resources, slow training, and undermine model improvements. These challenges have motivated a growing body of techniques for LLM RLOps, which manages the efficiency and reliability of operations throughout the LLM reinforcement learning training lifecycle. Despite growing interest, these techniques remain scattered across research areas, with limited synthesis of their operational roles and interactions. To bridge this gap, we present a systematic survey of LLM RLOps methods. We organize existing approaches into five stages: Training Setup, Experience Collection, Reward Computation, Policy Optimization, and Model Evaluation. Within each stage, we examine operational concerns, management techniques, and evaluation evidence. Building on this taxonomy, we identify open challenges and future directions in training process observability, cross-stage coordination, generalizable RLOps, multi-run management, and RLOps for recursive self-improvement. These directions extend operational management from individual training runs to diverse training settings, shared infrastructure, and successive improvement cycles. We maintain a GitHub repository indexing relevant papers and open resources at https://github.com/zhanglingzhe0820/Awesome-RLOps.

cs.SE · cs.CL · cs.CR

新论文 · v1arXiv:2611.22665

Contract-Explicit Evaluation: From Auditing Learned Readers to Scoped Certificates

Hao Lu, Weixing Zhu, Ruiyang Li

原文摘要

Evaluation readings of document-level relation extraction (DocRE) systems depend jointly on the model and the evaluation contract: gold version, scoring function, threshold. Rankings reverse across contracts, yet no existing method determines which comparisons survive the reversal. We address this gap with contract-explicit evaluation: a preregistered five-step protocol (Declare, Measure, Scope, Certify, Hedge), operationalizing the classical concordance decomposition of Kendall's rank correlation, that scopes certifiable regions and delivers distribution-free census certificates across two ecosystems. In the DocRE audit (44 training runs of one in-house architecture family), the scoring dimension scopes to a census certificate (three flagged instances, Clopper--Pearson upper bound 0.146); the confirmatory record includes registered negatives, and the audit prices its own simplification finding---the calibration dial's entire margin is one withdrawn commit. In an LLM-judge census (17 judges, 49,938 battles), 12.25% of leaderboard comparisons require hedging at the pre-declared split criterion (bound 0.124; panel-robust 0.143), with an exploratory two-tier reading certifying 69% at a bound \le 0.072 and a validity decomposition against released human labels (65% of contested battles are judge error, not ambiguity). The cross-model certificate is withdrawn on real confidence scores; the failure spectrum maps the boundary. The protocol, map, ledger, and per-instance artifacts are released.

cs.CL

新论文 · v1arXiv:2611.22541

From Perception to Simulation: Unveiling and Empowering Mental Spatial Dynamics in MLLMs

Jiachun Li, Zhuoran Jin, Kang Liu, Jun Zhao

原文摘要

Spatial reasoning is a fundamental component of world understanding for Multimodal Large Language Models (MLLMs). While existing research has advanced the perception of spatial relations in static scenes, the domain of dynamic spatial simulation remains relatively unexplored. To bridge this gap, we introduce MentalGrid, a benchmark focused on spatial reasoning tasks that require models to simulate evolving spatial states without intermediate visual feedback. Extensive evaluation on our benchmark reveals a significant deficiency in the spatial simulation capabilities of existing models. Specifically, we find that performance is constrained by perceptual limitations in simple simulation scenarios and hindered by inadequate multi-step reasoning as complexity increases. To address these challenges, we propose a three-stage curriculum learning framework utilizing 16k specialized spatial reasoning samples. This approach yields SpatialWalker, which achieves a 33.8% average improvement over its base model, surpassing proprietary models like Gemini-3-Flash in spatial simulation. Furthermore, SpatialWalker demonstrates robust generalization, maintaining a 4.9% average gain on out-of-domain benchmarks, demonstrating the pivotal role of spatial simulation as a fundamental building block for broader spatial intelligence.

cs.CV · cs.AI · cs.CL

新论文 · v1arXiv:2611.21019

MemDoctor: Stage-wise Error Attribution and Repair for MLLM Agent Memory

Ruiyang Jiang, Xinyang Liao, Weikai Huang, Yuqi Niu, Junjie Zhou, Yubin Zheng

原文摘要

Multimodal large language models (MLLMs) now assist users across sessions of images and text, carried by an external memory of what was said and shown. When such an assistant answers wrongly, the failure usually began earlier, in how a record was written, updated or retrieved, and the answer alone does not say where. We formalize multimodal memory as four stages, Write, Update, Retrieve and Answer, and define nine error types across them. A stage-resolved analysis on Mem-Gallery traces most failures to a source stage, and most of those lie before the answer. Injecting known errors against fact-preserving control edits, we then follow each tainted record downstream. Early errors survive to the answer as apparent evidence, and hurt most when no other record holds the fact. A diagnosis that trusts the quoted evidence therefore blames the stage where the error surfaces, not the one where it began. We propose MemDoctor, in which the MLLM that built the memory diagnoses its own failures. It traces each needed fact backward through the four stages, accepts a diagnosis only when the cited evidence is found in memory, and repairs that stage before answering again. On a 4B backbone, MemDoctor improves accuracy on both Mem-Gallery and MemLens, outperforming all evaluated 8B baselines, and the same loop also improves MuRAG and A-Mem.

cs.CL

新论文 · v1arXiv:2611.20891

From Recursion to Convolution: Parallel Self-Improvement for Multi-Agent Systems

Zheng Wang, Yuang Liu, Sachith Abeywickrama, Yangkai Ding, Tao Yu, Xinyu Zhang, Hongbo Wang, Qi Ye 等 10 位作者

原文摘要

Multi-agent self-improvement enables agents to improve through interaction and feedback. However, sequential dependencies across agents and refinement rounds can propagate errors, increase computational costs, and limit parallelism. Moreover, refining individual features or agents in isolation can overlook their dependencies and restrict collective coordination. To address these challenges, we propose ConvRSI, a framework for parallelizing multi-agent self-improvement through convolution. ConvRSI couples intra-agent convolution, which jointly refines related agent features, with inter-agent graph convolution, which aggregates messages according to the collaboration topology. Both stages operate in latent space, enabling parallel updates without repeatedly decoding and re-encoding intermediate messages. Experiments on MATH-500 show that ConvRSI achieves 5.04% performance improvement while reducing 17.27% inference cost, and 45.11% latency compared with RecursiveMAS. It also supports diverse collaboration patterns by representing their topologies as adjacency matrices. These results demonstrate the potential of convolutional co-evolution for efficient multi-agent self-improvement. Our code and datasets are provided in https://github.com/LYA2333/ConvRSI.

cs.AI · cs.CL · cs.LG

版本更新 · v2arXiv:2609.11900

MindTopo: Can Foundation Models Reason in Topological Space?

Yunfei Ge, Anbang Liu, Qineng Wang, Johnalbert Garnica, Jianwen Lyu, Zihan Wang, Reuben Tan, Jianfeng Gao 等 12 位作者

原文摘要

Spatial reasoning depends not only on metric properties such as distance, angle, and shape, but also on topological relations that remain invariant under continuous deformation. Cognitive science identifies these relations as foundational to spatial understanding, yet foundation-model evaluations largely focus on metric or viewpoint-dependent relations. We introduce MindTopo, a benchmark of topological intuition across five properties grounded in cognitive science and formal topology, namely continuity, separation, order, enclosure, and knots. MindTopo evaluates each property at two cognitive levels. Reasoning asks a model to identify topological relations or infer how they change. Planning instantiates a foundation model as a closed-loop agent whose policy selects environment actions. MindTopo provides a Core Set of 3,900 instances across 13 procedurally generated task types, split into an Easy and a Hard level, and an Extended Set of 11,035 original instances. On the Core Set, we benchmark 13 multimodal large language models (MLLMs) and study agent configurations augmented with image and video generation, covering seven video generative models in planning settings. Every MLLM performs better on reasoning than on planning. On the Easy level, where human scores are available, the strongest model trails humans in both reasoning and planning. On Qwen3-VL-2B-Instruct, supervised fine-tuning and reinforcement learning improve reasoning more than planning. Generated observations retain local cues and reach solved endpoints, yet none of the 1,890 evaluated video rollouts passes all process-validity checks.

cs.AI · cs.CL · cs.CV

版本更新 · v2arXiv:2601.02236

CD4LM: Consistency Distillation and aDaptive Decoding for Diffusion Language Models

Yihao Liang, Ze Wang, Ximeng Sun, Jiang Liu, Jialian Wu, Hao Chen, Xiaodong Yu, Emad Barsoum 等 10 位作者

原文摘要

Diffusion language models (DLMs) accelerate generation by decoding multiple tokens in parallel, but pushing them to fast decoding often sacrifices accuracy. We attribute this to a static-to-dynamic mismatch: training optimizes local denoising under fixed schedules, while fast inference skips steps and encounters irregular mask patterns. Existing accelerators largely address only one side of this problem. Training-free inference methods speed up decoding without changing the model, leaving the mismatch unresolved; training-side distillation methods adapt the model, but often tie the student to a specific decoding trajectory, limiting composability and still suffering accuracy loss. We propose CD4LM, a composable framework for accurate DLM acceleration. Its core training component, Discrete-Space Consistency Distillation (DSCD) trains a trajectory-invariant student on diverse masked states, reducing the mismatch without fixing a decoding policy. We pair DSCD with a lightweight ConfidenceAdaptive Decoder, while keeping the trained student compatible with external inference-side accelerators. On GSM8K, CD4LM preserves accuracy while achieving a 5.18x speedup; stacking CD4LM with per-step cache accelerators further pushes speedup to 9.73x at teacher-level accuracy, suggesting a scalable path for combining training-side robustness with future decoding advances. Code is available at https://github.com/yihao-liang/CDLM.

cs.CL

新论文 · v1arXiv:2611.19800

Multi-Branch Transformer

Gaotang Li, Lingjie Chen, Chunyuan Deng, Zhiwei Xu, Hongzheng Chen, Christopher De Sa, Jingrui He, Yingheng Wang 等 9 位作者

原文摘要

Transformer language models are commonly scaled by making their blocks wider or stacking them deeper. In this work, we explore a complementary path: branching, the parallel composition of full Transformer blocks. We introduce the Multi-Branch Transformer (MBT), which processes the input through shared initial layers, executes independently parameterized branches of sequential blocks in parallel, and fuses their outputs before shared final layers. At comparable training compute, we study three MBT variants: Wide-Shallow preserves block width, Narrow-Deep preserves sequential depth, and Balanced interpolates between them. In controlled pretraining experiments at four mixture-of-experts scales (0.8B to 2.1B total parameters), 2-branch Wide-Shallow MBTs consistently outperform sequential Transformers under matched parameter and training-compute budgets. Balanced MBTs improve further and retain comparable quality as the number of branches increases. To improve inference efficiency at long contexts, we introduce Joint-KV, which constructs a shared KV cache from branch representations at each depth. Our final recipe combines Balanced MBTs with Joint-KV, supported by an inference engine that exploits branch grouping and shared-cache reuse. Fitted scaling curves indicate that standard Transformers require up to 1.26x as much training body compute to match our final models' validation loss. These models reduce full-model KV-cache storage by up to 34% and deliver up to 51.4% higher decoding throughput on a single H200 GPU. Ablation studies show that branching yields gains beyond changes in width and depth, enables an effective KV-sharing strategy (Joint-KV), remains robust across fusion functions, and performs best in the middle layers. These findings show that branching offers a practical path for scaling language models toward both better modeling quality and faster inference.

cs.LG · cs.CL

版本更新 · v2arXiv:2611.17332

ArchSpace:记录、评审与引用 LLM 架构研究

ArchSpace: Recording, Reviewing, and Referencing LLM Architecture Research

Jie Wang, Tong Zhu, Jiahao Kuang, Zhe Xu, Mowen Ruan, Yunkai Chai, Yuliang Liu, Dehong Zheng 等 14 位作者

AI 速读 基于摘要

ArchSpace 是一个公开平台,用于记录 LLM 架构提案及其实验与推理、由社区评审,并供后续工作引用,同时借助缩放定律在小规模实验上预测更大规模的模型损失。

cs.CL · cs.LG

查看原文摘要

Model architecture shapes the scaling behavior, training efficiency, and inference cost of LLMs, making it a central focus of both industry and academia. In industry, teams such as those behind DeepSeek, Qwen, and Kimi routinely introduce architectural innovations, yet public reports rarely document which ideas succeeded, which were abandoned, and why. In academia, limited compute and inconsistent baselines often restrict evaluations to small-scale, single-point comparisons, making it difficult to establish whether improvements hold at larger scales or translate into broader impact. ArchSpace addresses both gaps as a public platform for Recording, Reviewing, and Referencing Architecture Research: each proposal is recorded with its experiments and reasoning, reviewed by the community, and referenced by later work. To evaluate ideas before costly large-scale training, ArchSpace leverages scaling laws to predict model loss at larger scales from smaller experiments, a practice widely adopted in industry. We provide initial references that make the platform usable: Qwen3.5 and OLMo3 baselines, fitted scaling laws, and a reusable training recipe. Ideas validated through this review can in turn become new baselines, and ArchSpace thereby aims to broaden the range of ideas that shape the next generation of models.

版本更新 · v2arXiv:2601.10455

Beyond Trajectory Similarity: Evaluating Surgical Planning by Protocol Adherence

Ruochen Li, Kun Yuan, Yufei Xia, Yue Zhou, Qingyu Lu, Weihang Li, Youxiang Zhu, Nassir Navab

原文摘要

Surgical planning is commonly evaluated by trajectory similarity to a single observed execution, a paradigm borrowed from activity anticipation benchmarks such as Ego4D. This conflates two distinct concepts: task completion and protocol adherence and assumes similarity to a completed surgical trajectory implies a valid plan. In surgery, this assumption is flawed: the clinically relevant target is protocol adherence rather than mere completion, and the reference trajectory is only one observed execution that may itself deviate from the protocol. We therefore introduce an evaluation framework grounded in expert-defined protocol rules that directly measures protocol adherence. Under this framework, similarity-based metrics systematically fail: they penalize protocol-adherent plans that deviate from the reference trajectory, while rewarding protocol-violating plans that remain superficially similar to it. Evaluating Video-LLMs further reveals two separable failure modes: unreliable video-to-state grounding and under-constrained planning. Neither is resolved by model scaling alone, whereas structural knowledge injection provides consistent gains. Our results motivate a shift from trajectory similarity toward protocol adherence as the evaluation standard for surgical planning.

cs.CL · cs.RO

新论文 · v1arXiv:2611.18096

VEPO: Mitigating Multilingual Hallucinations in LVLMs via Visual Evidence-guided Preference Optimization

Chengpeng Fu, Xiaocheng Feng, Wenshuai Huo, Yichong Huang, Baohang Li, Duyu Tang, Ting Liu

原文摘要

Large vision-language models (LVLMs) achieve strong multimodal performance, yet they remain prone to object hallucination---generating answers that are linguistically fluent but visually ungrounded. This issue is especially severe in low-resource languages, where weaker cross-modal alignment causes models to rely more on language priors than on image evidence. Existing remedies either require costly multilingual annotations or external translation/reward models, or intervene at inference time in ways that can distort the output language. We propose VEPO(Visual Evidence-guided Preference Optimization), a training-time method that transfers reliable visual grounding from English to other languages without external models. VEPO first localizes key visual evidence using English queries via mid-layer cross-modal attention, then constructs preference pairs that share the same correct answer but differ in evidence sufficiency (full image vs.\ key-region masked image). Direct preference optimization on these pairs encourages low-resource inputs to prefer answers supported by complete visual evidence. Experiments on multilingual POPE and MME with Chinese, French, and Hindi show consistent hallucination reduction and perception gains, with the largest improvements on the lowest-resource language. Analyses further confirm that VEPO sharpens question-relevant attention and that evidence-guided masking outperforms random masking.

cs.CL

版本更新 · v3arXiv:2510.08697

BigCodeArena: Unveiling More Reliable Human Preferences in Code Generation via Execution

Terry Yue Zhuo, Xiaolong Jin, Hange Liu, Juyong Jiang, Tianyang Liu, Chen Gong, Bhupesh Bishnoi, Vaisakhi Mishra 等 40 位作者

原文摘要

Crowdsourced model evaluation platforms, such as Chatbot Arena, enable real-time evaluation from human perspectives to assess the quality of model responses. In the coding domain, manually examining the quality of LLM-generated content is extremely challenging, as it requires understanding long chunks of raw code and deliberately simulating code execution. To this end, we introduce BigCodeArena, an open human evaluation platform for code generation backed by a comprehensive and on-the-fly execution environment. Built on top of Chatbot Arena, BigCodeArena enables the execution of LLM-generated code and allows humans to interact with the execution process and outcomes. We collected over 14,000 raw code-centric conversation sessions across 10 widely used LLMs, spanning 10 languages and 8 types of execution environments. Among these conversations, we identified more than 4,700 multi-turn samples with pairwise human preferences. Further analysis uncovers underexplored preferences of LLMs in fine-grained domains characterized by tasks, languages, and frameworks. To systematically examine code understanding and generation capabilities of frontier LLMs, we curated two benchmarks based on the collected data, namely BigCodeReward and AutoCodeArena. For BigCodeReward, we post-processed the 4,700 conversations and evaluated the consistency between reward models and human preferences. The evaluation shows that most LLMs have superior performance in judging coding preferences when the execution results are available. Inspired by these findings, we propose AutoCodeArena, an automatic Elo rating benchmark designed to assess the coding quality of LLMs without human involvement. We find that proprietary LLMs like GPT-5, Claude-Sonnet-4, and Claude-Opus-4 still lead in code generation performance among recent emerging models.

cs.SE · cs.AI · cs.CL

版本更新 · v4arXiv:2301.12868

On Robustness of Prompt-based Semantic Parsing with Large Pre-trained Language Model: An Empirical Study on Codex

Terry Yue Zhuo, Zhuang Li, Yujin Huang, Fatemeh Shiri, Weiqing Wang, Gholamreza Haffari, Yuan-Fang Li

原文摘要

Semantic parsing is a technique aimed at constructing a structured representation of the meaning of a natural-language question. Recent advancements in few-shot language models trained on code have demonstrated superior performance in generating these representations compared to traditional unimodal language models, which are trained on downstream tasks. Despite these advancements, existing fine-tuned neural semantic parsers are susceptible to adversarial attacks on natural-language inputs. While it has been established that the robustness of smaller semantic parsers can be enhanced through adversarial training, this approach is not feasible for large language models in real-world scenarios, as it requires both substantial computational resources and expensive human annotation on in-domain semantic parsing data. This paper presents the first empirical study on the adversarial robustness of a large prompt-based language model of code, \codex. Our results demonstrate that the state-of-the-art (SOTA) code-language models are vulnerable to carefully crafted adversarial examples. To address this challenge, we propose methods for improving robustness without the need for significant amounts of labeled data or heavy computational resources.

cs.CL

版本更新 · v2arXiv:2606.31272

The Decomposition Is the Fingerprint: Per-Component Identity for Agent Skills

Hongliang Liu, Yuhao Wu, Tung-Ling Li

原文摘要

AI agents increasingly acquire and execute skills at runtime: bundles of prompt instructions, executable code, and tool declarations fetched from marketplaces and other agents. Governing them needs a stable notion of skill identity, yet cryptographic hashing is engineered to destroy the very similarity we need, as a one-character edit scrambles the digest. We present a compact, locality-sensitive fingerprint that embeds each component of a skill and projects it to bits with a multi-bank SimHash, giving a fixed 120-byte signature compared in constant time by Hamming distance. Our central claim is that keeping the fingerprint as a per-component triple (prompt, code, tools), rather than a single score, is what makes it useful: the triple recovers skill-family identity through paraphrase, renaming, refactoring, and controlled code translation when another component remains shared, while independent multilingual reimplementation is not recovered; it also localizes which component carries the reuse. We claim lineage, not behavioral equivalence: identity supplies the structural axis of a registry and leaves safety to behavioral verification. The fingerprint reaches an area under the ROC curve (AUC) of 0.974 (95% CI [0.956, 0.994]) over 4,950 pairwise comparisons while using 77x fewer bits than the embedding it approximates, with ranking preserved in expectation and finite-bit concentration; the per-component split turns one number into relationship classification, families, novelty, and a portable "SkillBOM" for a skill registry. On a 906-skill injection benchmark the fingerprint recognizes injected skills as tampered copies of a known base and localizes the change, but recognition is not trust: it remains, by design, an identity signal complementary to behavioral verification rather than a safety verdict.

cs.CR · cs.CL · cs.LG