【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 www.xiaoyuzhoufm.com
【目录】
本期的 15 篇论文如下:
00:27 🎮 AgentGarten: Code Worlds for Evolving Agents(AgentGarten:面向进化智能体的代码世界)
01:03 🎮 Learn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments?(Learn2Play Bench:LLM 智能体在陌生环境中从经验中学习的效果如何?)
01:43 🚦 TokenRouter: Efficient Serving System for Token-Level LLM Routing(TokenRouter:面向 Token 级 LLM 路由的高效服务系统)
02:29 🌍 From Traces to Agentic Worlds: Agentic Language World Models for Interactive Environment Simulation(从轨迹到智能体世界:用于交互式环境模拟的智能体语言世界模型)
03:26 🤖 SuperNav: An Agentic Navigation System for Any Task in Any Scene(SuperNav:面向任意场景中任意任务的智能体导航系统)
04:11 🚀 MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement(MiMo-V2.6:将强化学习扩展至自我改进)
04:55 🤖 In-context Robot Learning Made Simple: A Democratized Recipe for Manipulation Tasks(让机器人上下文学习变简单:面向操作任务的普惠方案)
05:39 🤖 Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction(面向细粒度具身交互的多智能体自我中心世界模型)
06:27 🤖 DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training(DreamTrue:基于反事实后训练的动作忠实机器人世界模型)
07:05 🌀 OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D Cinemagraphs(OuroWorld:将任意3D世界变为多样且无限循环的3D动态影像)
07:50 ⚡ MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers(MC-Sparse:解构并弥合扩散 Transformer 中的稠密-稀疏注意力差距)
08:34 🔗 Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching(超越时空先验:面向稠密对应匹配的可泛化方法)
09:16 🧪 TestPrism: Rethinking Test Evaluation Beyond a Single Reference(TestPrism:重新思考超越单一参考的测试评估)
10:02 🎨 Post-Training Frontier Text-to-Image Models by Composing Preference and Rubric Rewards(通过组合偏好奖励与评分规则奖励对前沿文本到图像模型进行后训练)
10:51 ✂ SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference(SparseDecoding:面向准确高效大语言模型推理的解码感知剪枝)

【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
