👍 72
05/30 08:00
Recent progress in the development of language models has been defined by scale, with each generation absorbing more of the world's knowledge into its weights. However, many practical applications benefit more from robust reasoning than from extensive parametric knowledge. In this setting, task-spec
中文介绍 针对大语言模型过度依赖参数知识的问题,提出OCC-RAG框架,通过检索增强实现忠实问答,强调稳健推理而非规模扩展,在知识密集型任务上提升可靠性。
👍 40
05/22 08:00
Identifying which brain regions represent a visual concept in the human brain is a central challenge in neuroscience. Existing approaches have localized coarse functional regions (e.g., faces, places) through activation maximization, identifying regions that activate strongly for a target concept re
中文介绍 提出从激活到因果的框架,通过因果推断揭示人脑中视觉概念的表征区域,超越传统粗粒度功能定位,实现更精细的脑区功能映射。
👍 36
06/01 08:00
Search agents are often trained as policies over growing transcripts: the model must decide how to search while also remembering what it has seen, which evidence is useful, which constraints remain open, and which claims have actually been checked. We argue that this formulation puts too much routin
中文介绍 针对搜索代理在增长转录中决策与记忆耦合导致的效率问题,提出Harness-1框架,通过状态外化将搜索决策与证据记忆分离,降低策略学习难度,提升搜索效率。
👍 33
05/31 08:00
On-Policy Distillation (OPD) is a fundamental technique for efficient post-training of large language models (LLMs), with broad applications in agent learning, multi-task enhancement, and model compression. However, OPD training becomes unstable when the teacher and student distributions differ subs
中文介绍 针对on-policy蒸馏中师生分布差异导致训练不稳定的问题,提出信任区域策略蒸馏(TR-OPD),通过限制策略更新幅度确保训练稳定性,适用于LLM后训练、智能体学习与模型压缩。
👍 26
06/02 08:00
Test-time scaling is a powerful approach to obtain better reasoning in large language models, but it becomes memory-bottlenecked during long-horizon decoding, as the KV-cache grows. KV-cache quantization can help improve this, but current methods are evaluated under prefill-like settings and errors
中文介绍 针对长推理中KV缓存量化误差累积的问题,提出KVarN方差归一化量化方案,通过对KV缓存进行方差归一化降低量化误差,在保持推理性能的同时减少内存占用。
👍 24
06/01 08:00
Reinforcement learning (RL) post-training improves large language models (LLMs) on individual domains such as mathematical reasoning, code generation, question answering, and creative writing (CW), but training on one domain often degrades performance on others. Existing explanations based on catast
中文介绍 针对多领域RL后训练中跨领域干扰与性能恢复问题,提出局部扰动理论框架,建模能力迁移与遗忘机制,为多任务学习中的领域冲突提供理论基础。
👍 21
06/02 21:07
World models and multimodal large language models (MLLMs) provide complementary capabilities for predicting future outcomes from static visual observations. World models can generate concrete visual rollouts of possible futures, while MLLMs can reason abstractly over questions, goals, and rules. How
中文介绍 探索世界模型与多模态大语言模型(MLLM)在具体视觉滚动与抽象推理上的互补性,提出结合两者优势的框架,实现对视觉观测的未来预测与高阶推理。
👍 20
06/01 08:00
Autonomous agents are increasingly expected to support end-to-end medical-AI research workflows, moving beyond isolated prediction tasks or short-form clinical question answering. However, existing medical agent benchmarks primarily evaluate final outputs, providing limited visibility into agent beh
中文介绍 针对医疗AI代理基准缺乏过程可见性的问题,提出AutoMedBench,评估自主代理在完整医疗研究流程中的表现,涵盖多步骤任务与中间推理。
👍 19
05/29 08:00
Mid-training has become an important stage in modern LLM development, using large-scale curated mixtures to strengthen capabilities before final post-training. Its data selection problem is distinct: the data are optimized under a pretraining-style objective at near-pretraining scale, but are curate
中文介绍 针对LLM中期训练数据选择问题,提出MIRA(规则锚定源感知数据选择),利用评分规则和大规模预训练数据筛选策略,提升中期训练阶段的能力强化效果。
👍 16
06/01 10:52
Reinforcement learning (RL) for visual reasoning needs scalable, verifiable, and controllable training signals. Existing visual RL post-training trains on static curated datasets, with fixed image-question-answer samples bounded by their collection budget. In this work, we introduce TRON (Targeted,
中文介绍 针对视觉推理强化学习训练信号不足的问题,提出TRON在线环境,通过规则可验证的目标生成实现可扩展训练信号,突破静态数据集限制。
👍 15
06/02 08:00
Understanding a video requires more than recognizing isolated moments, as humans continuously track entities, states, and events over time. This capacity for visual state tracking is fundamental to video understanding, yet remains underexplored in current evaluations of Multimodal Large Language Mod
中文介绍 针对MLLM视频理解中实体、状态与事件的连续跟踪能力评估缺失,构建视觉状态跟踪基准,系统衡量模型对动态场景的时序理解能力。
👍 14
06/02 08:00
The past few decades have witnessed significant advances in the design of machine learning algorithms, from early studies on task-specific shallow models to more general deep Large Language Models (LLMs). Despite showing promising results in tasks that require instant prediction or in-context learni
中文介绍 借鉴生物睡眠机制,提出让语言模型通过「睡眠」阶段进行自我修改与记忆巩固,提升长期记忆保持与知识整合能力,缓解灾难性遗忘。
👍 13
06/02 08:00
As autonomous vehicle capabilities advance, the safe evaluation of driving policies in long-tail scenarios remains a critical bottleneck. In closed-loop simulation, the driving policy model actively interacts with the environment, where its actions dynamically update the simulator state and directly
中文介绍 针对自动驾驶闭环仿真中长尾场景评估瓶颈,提出OmniDreams实时生成式世界模型,动态响应驾驶策略行为,实现安全高效的闭环模拟。
👍 12
06/02 08:00
Modern generative models possess a deep understanding of visual content, yet training them for image editing typically requires massive datasets of paired examples. This limits scalability, especially for video editing where collecting paired data is prohibitively expensive. We propose Bootstrap You
中文介绍 针对图像编辑依赖配对数据的局限,提出基于流匹配的自举生成器方法,无需配对数据即可训练编辑模型,并扩展至视频编辑,显著降低数据收集成本。
👍 11
06/02 02:20
Personalization is a crucial capability of modern language agents. However, current research primarily positions personalized agents as passive responders to user preferences, limiting their ability to interact with users and provide suggestions or guidance proactively. To systematically evaluate su
中文介绍 针对语言代理在说服性对话中主动影响用户的能力评估缺失,提出Ψ-Bench基准,系统测试代理感知用户人格并调整说服策略的效果。
👍 10
06/01 14:51
Auto-harness systems such as A-Evolve, GEPA, and Meta-Harness improve LLM agents by optimizing prompts, skills, tools, memories, and supporting infrastructure from execution feedback, but they are typically evaluated on fixed offline benchmarks. Real deployments instead present open-ended task strea
中文介绍 针对开放任务流中LLM代理部署的持续改进需求,提出自适应自动马甲系统,通过执行反馈动态优化提示、技能与工具,实现在线持续自我提升。
👍 10
06/02 11:42
Test-time scaling improves the reasoning performance of large language models but incurs substantial cost in both total computation and latency. Existing adaptive sampling methods partially mitigate this issue by dynamically deciding when to stop sampling, yet they typically rely on heuristic rules
中文介绍 针对LLM测试时扩展计算与延迟成本高的问题,提出小RL控制器引导的自适应采样策略,动态决定停止采样时机,在保证推理质量的同时降低总计算量。
👍 10
06/01 08:00
Reinforcement learning (RL) improves large language model (LLM) agents by teaching them which actions lead to high rewards, but provides little supervision on what those actions do to the environment. World modeling (WM) can fill this gap, yet existing approaches often require separate simulators, e
中文介绍 针对语言代理RL训练缺乏环境模型的问题,提出策略与世界模型协同训练框架,联合学习动作选择与环境动态,提升代理的规划能力与样本效率。
👍 8
06/02 08:00
We introduce PaddleOCR-VL-1.6, an upgraded compact document parsing model built upon PaddleOCR-VL-1.5. Although PaddleOCR-VL-1.5 establishes a strong 0.9B baseline, its remaining errors concentrate in under-optimized regions where model behavior is unstable, data coverage is sparse, or supervision i
中文介绍 在PaddleOCR-VL-1.5(0.9B)基础上,提出PaddleOCR-VL-1.6,通过欠优化区域细化和渐进式后训练策略,重点修复不稳定、缺乏监督和稀疏数据区域的错误,提升文档解析精度。
👍 8
05/28 08:00
Long chain-of-thought (CoT) traces are widely used as supervision for reasoning-oriented LLM SFT, yet answer-correct traces can still lead to markedly different fine-tuning outcomes. We study post-conclusion continuation in answer-correct long-CoT data: a continuation where the answer appears suffic
中文介绍 针对答案正确的长思维链追踪中存在的后结论有害续写问题,系统诊断该现象对微调效果的影响,提出过滤策略以提升长CoT训练数据的质量。