👍 101
09/26 08:00
Reinforcement learning (RL) for code agents often uses executable tests to provide binary rewards. With these rewards, Group Relative Policy Optimization (GRPO) assigns identical advantages to test-passing trajectories within each rollout group, overlooking differences in implementation quality and
中文介绍 针对代码 Agent 强化学习,提出基于分组 Agentic 评分与优势再分配的改进:超越仅用可执行测试的二元奖励,对组内通过的轨迹按实现质量差异进行评分,并在 GRPO 框架下重新分配 advantage,提升对推理与实现质量的区分度。
👍 55
09/28 08:00
Most modern multimodal large language models (MLLMs) build on a pretrained visual encoder that provides a strong visual prior. Encoder-free MLLMs instead learn visual representations directly from raw pixels, offering a simple and unified architecture, but their scaling behavior has not been systema
中文介绍 研究去除视觉编码器的 Encoder-Free 多模态预训练,通过 scaling laws 系统分析其从原始像素直接学习视觉表征的行为与规模效应,评估能否在简洁统一架构下替代传统预训练 visual encoder,为 MLLM 架构选择提供依据。
👍 30
09/27 08:00
A central goal in robotics is to enable manipulation across changing tasks and environments. Vision-language-action (VLA) models provide broad manipulation capabilities but can struggle when execution requires diagnosing failures and adapting behavior. Strong agents can discover effective interventi
中文介绍 提出跨 Agent 的递归 Harness 蒸馏用于机器人操作:让强 Agent 诊断失败并发现有效干预,再把这些干预递归蒸馏到 VLA 模型中,提升其在变化任务与环境下的操作能力,减少执行失败。
👍 27
09/27 08:00
Large language model (LLM) agents increasingly undertake extreme-long (xlong) horizon tasks, where a single execution can span hours, hundreds of model--environment interactions, and nearly 1M tokens per rollout. Applying online reinforcement learning (RL) to such executions poses two fundamental ch
中文介绍 提出 QwenGyre 弹性强化学习框架,用于训练 xlong-horizon Agent。针对单次执行跨越数小时、数十万至近百万 token 的 rollout,解决在线 RL 面临的计算与资源挑战,支持长时程交互任务的稳定训练。
👍 25
09/28 08:00
Vision-language (VL) pretraining using paired chest X-ray (CXR) images and radiology reports has shown strong potential for medical image understanding. However, existing methods often remain dependent on task-specific finetuning because radiology reports are lengthy, clinically dense, and difficult
中文介绍 SentZero 提出以句子为中心增强的视觉-语言预训练,用于多任务零样本胸部 X 光分析。通过解析临床密集且冗长的放射学报告,减少对任务特定微调的依赖,提升医疗图像理解的零样本迁移能力。
👍 17
09/26 08:00
Fast matrix multiplication algorithms keep the product fixed and search for a cheaper way to evaluate it. We instead ask whether a Transformer's learned projections can use a different, cheaper product altogether. Building on an associative-algebra construction that replaces ordinary matrix multipli
中文介绍 提出 Associative Algebra Layers 用于 Transformer:不再保持矩阵乘积不变而优化计算,而是允许学习到的投影使用不同但更廉价的乘法结构,基于结合代数构造降低计算成本,同时保持参数规模不变。
👍 16
09/27 08:00
Coding-agent evaluation has progressed from resolving individual issues to carrying out long-horizon development, yet task completion is still largely assessed within a single codebase. In software ecosystems, many features and bug fixes require coordinated changes across multiple repositories. We i
中文介绍 提出 WideSWE 基准,评估编码 Agent 能否跨多个仓库协调修改。突破单代码库内完成任务的局限,衡量在真实软件生态中同步多仓库特性的能力,揭示多仓库协作的新挑战。
👍 15
09/28 08:00
Frontier general-purpose systems are rapidly expanding beyond visual understanding into capabilities traditionally handled by dedicated computer-vision models. As these capabilities expand, a central question for the computer-vision community is how far this reach extends, and what remains hard. We
中文介绍 评估 GPT-6 Astra 在计算机视觉任务上的泛化能力,区分「Hard Vision」与「Easy Vision」。系统分析前沿通用模型对传统视觉专用模型的替代范围,揭示仍具挑战的视觉任务边界。
👍 15
09/28 08:00
Tool-based image editing (image retouching) is commonly formulated with autoregressive multimodal large language models (MLLMs) that sequentially generate reasoning, tool selections, and parameter values. In this work, we present a novel approach to tool-based image editing by framing the task as a
中文介绍 FlowTool 将基于工具的图片修图任务重构为 flow matching 问题,而非依赖自回归 MLLM 顺序生成推理、工具选择和参数。通过 flow matching 控制工具参数,提升图像编辑的生成质量与可控性。
👍 11
09/28 08:00
As language models take a growing role in AI development, a natural aspiration is for them to reflect on their own learning process, as humans do, and use that reflection to improve themselves. At the same time, these models have an advantage that human learners lack, since training leaves parameter
中文介绍 Imprint Reader 提出从权重更新读出到行为干预的机制,使语言模型反思自身学习过程并利用其进行自我改进。利用训练留下的参数痕迹作为线索,推动模型自省与自主优化。
👍 10
09/28 08:00
We study on-policy distillation (OPD) through the lens of reinforcement learning, establishing a connection between the reverse-KL objective in OPD and KL-regularized policy optimization. Building on this connection, we introduce Least-Square Policy Distillation (LSPD), an RL-inspired framework that
中文介绍 从强化学习视角研究 on-policy distillation,建立 OPD 中 reverse-KL 目标与 KL 正则策略优化之间的联系。提出 Least-Square Policy Distillation,提升 LLM 推理的样本效率与蒸馏质量。
👍 9
09/28 08:00
Reinforcement learning (RL) post-training for large language models (LLMs) coordinates multiple models across generation, inference, and training on GPU clusters. Several factors may change during a run, including resource availability, sequence length, memory pressure, and stage bottlenecks. As a c
中文介绍 Nereus 提出用于 LLM 后训练的自适应并行方案,针对 RL 训练中 generation、inference、training 多阶段在 GPU 集群上的协调问题。根据资源可用性、序列长度、显存压力与瓶颈阶段动态调整并行策略。
👍 8
09/26 08:00
Learning from experience in LLM agents has become a key paradigm for developing self-evolving agents that continuously learn and expand their capabilities. Within this paradigm, synthesizing the agent skill has emerged as a promising solution for transforming accumulated experience into reusable pro
中文介绍 ExpVoyager 提出面向动态 Agent 技能合成的直接经验导航,用于自演化 LLM Agent。通过将累积经验转化为可复用技能,提升 Agent 的持续学习与能力扩展效率。
👍 6
09/27 08:00
Verified solutions are not equally useful for preparing reasoning models for reinforcement learning (RL). We present a comprehensive study of route diversity, the variation in the sequences of reasoning steps in supervised fine-tuning (SFT) data, and propose a lightweight, rule-based fingerprint to
中文介绍 研究表明 SFT 轨迹多样性对 Post-RL 泛化至关重要。提出 route diversity 概念,并用轻量规则式 fingerprint 筛选多样化 SFT 轨迹,提升 RL 后训练推理模型的泛化能力。
👍 5
09/28 08:00
We present OLIVE (OnLine InterVEntion). At each iteration, the evolving student policy generates a new prefix, the teacher continues it autoregressively, and the student is updated using cross-entropy computed on the teacher-generated tokens. Each design choice targets a corresponding limitation of
中文介绍 OLIVE(OnLine InterVEntion)让进化的学生策略生成前缀,教师再自回归续写,学生以教师生成 token 的交叉熵更新。针对既有方法的局限,提升学生模型从教师续写中学习的效率与适应性。
👍 5
09/26 08:00
Multiple agents may often conflict in an organization: for example, one coding agent changes an interface in a repository, but another continues to develop on the old version where existing tests become stale. A conversation can resolve the episode, but when the participants change, what makes the l
中文介绍 Relic 将多 Agent 协作转化为持久的组织能力。针对组织中 Agent 因接口变更、测试失效等问题产生的冲突,提出让协作成果沉淀为可持续复用的组织机制,避免参与者更替后协作断裂。
👍 4
09/28 08:00
RLVR provides reliable trajectory-level credit, while OPSD offers dense supervision for token-level credit. This exposes a fundamental coupling when updating step-level credit direction and magnitude with teacher supervision, preventing steps from receiving reliable credit directions and contributio
中文介绍 提出解耦 credit 方向与幅度的自蒸馏方法,针对 RLVR 轨迹级信号与 OPSD token 级监督耦合的问题。防止步骤获得不可靠的信用方向,提升自蒸馏中信用分配的稳定性与可靠性。
👍 4
09/28 08:00
Multi-vector retrievers built on vision-language models lead visual document retrieval (VDR), but they run a multi-billion-parameter query encoder on every search. Distilling this encoder into a small student that queries the teacher's existing index would remove the bottleneck. The standard recipe,
中文介绍 ColNanoVDR 通过最优传输实现免文档查询蒸馏,用于多向量视觉文档检索。将十亿级 query encoder 蒸馏进小型学生模型,复用教师已有索引,消除查询阶段算力瓶颈。
👍 4
09/28 08:00
Reliable AI safeguards require both control mechanisms that reduce unsafe behavior and monitoring mechanisms that detect safety risks during model interactions. Established behavioral safeguards include alignment methods that optimize model outputs and text monitors that assess interaction text. Rep
中文介绍 研究 Representation Engineering 在 LLM 安全中的作用,探讨模型内部表征何时对安全保障有帮助。覆盖控制机制与监控机制,评估表征级方法在降低不安全行为与检测风险中的效果。
👍 3
09/27 08:00
Connectionist temporal classification (CTC) naturally supports offline and streaming speech recognition with utterance-level supervision, but conventional implementations materialize frame-by-vocabulary activations in memory, making CTC training with native LLM vocabularies prohibitively memory-inte
中文介绍 提出 Pruned CTC,用于内存高效的大词汇 ASR 训练。传统 CTC 会将逐帧-逐词汇激活全部物化到显存,在大词表场景下不可行;该方法通过剪枝减少内存占用,支持原生 LLM 词表的 CTC 训练。