2026.06.12 | EvoArena追踪记忆演化，测试AI洞悉偏好变化

【赞助商】OpenClaw快报每天五分钟，听听 OpenClaw 快报，带你了解最新动态和业内讨论传送门 https://www.xiaoyuzhoufm.com/podcast/6a1732a2dffa135d0ab5ef43【目录】本期的 15 篇论文如下：[00:31] 🧠 EvoArena: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments（EvoArena：在动态环境中追踪记忆演化以实现鲁棒的LLM智能体）[01:32] 🧠 SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning（SpatialClaw：重新思考面向智能体空间推理的动作接口）[02:33] 🔍 FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents（FORT-Searcher：为训练深度搜索代理合成抗捷径搜索任务）[03:31] 🛠 Robust-U1: Can MLLMs Self-Recover Corrupted Visual Content for Robust Understanding?（鲁棒-U1：多模态大语言模型能否自我恢复受损视觉内容以实现鲁棒理解？）[04:27] 🔄 InterleaveThinker: Reinforcing Agentic Interleaved Generation（交错思考者：强化智能交错生成）[05:14] 🧮 MaxProof: Scaling Mathematical Proof with Generative-Verifier RL and Population-Level Test-Time Scaling（MaxProof：利用生成-验证强化学习与群体级测试时扩展实现数学证明的规模化）[06:11] 🧠 MiniMax Sparse Attention（MiniMax稀疏注意力机制）[06:50] 🖥 WeaveBench: A Long-Horizon, Real-World Benchmark for Computer-Use Agents with Hybrid Interfaces（WeaveBench：面向混合界面计算机使用代理的长期、真实世界基准）[07:50] 🔬 LabVLA: Grounding Vision-Language-Action Models in Scientific Laboratories（LabVLA：在科学实验室中落地视觉-语言-动作模型）[08:55] 🦾 HYDRA-X: Native Unified Multimodal Models with Holistic Visual Tokenizers（HYDRA-X：具备整体视觉分词器的原生统一多模态模型）[09:45] 🧩 N-GRPO: Embedding-Level Neighbor Mixing for Enhanced Policy Optimization（N-GRPO：基于嵌入级邻居混合的增强策略优化）[10:44] 🔬 EurekAgent: Agent Environment Engineering is All You Need For Autonomous Scientific Discovery（EurekAgent：智能体环境工程是自主科学发现所需的一切）[11:38] 🏃 VideoMDM: Towards 3D Human Motion Generation From 2D Supervision（VideoMDM：从二维监督迈向三维人体运动生成）[12:35] 🔍 Where, What, Why, and Importance: Structured Defect Grounding for Text-to-Image Feedback（位置、类型、原因与重要性：面向文本到图像反馈的结构化缺陷定位）[13:26] 🔀 Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning（揭秘隐藏状态循环：基于在线强化学习的可切换潜在推理）【关注我们】您还可以在以下平台找到我们，获得播客内容以外更多信息小红书: AI速递在小宇宙查看该单集文稿