【目录】
本期的 15 篇论文如下:
[] 🌌 Cosmos 3: Omnimodal World Models for Physical AI(宇宙3:面向物理AI的全模态世界模型)
[] 🎧 Audio Interaction Model(音频交互模型)
[] 🔍 Where Do Deep-Research Agents Go Wrong? Span-Level Error Localization in Agent Trajectories(深度研究型智能体错在哪里?智能体轨迹中的跨度级错误定位)
[] 🔍 Reproducing, Analyzing, and Detecting Reward Hacking in Rubric-Based Reinforcement Learning(在基于评分标准的强化学习中复现、分析与检测奖励作弊行为)
[] 🧭 OVO-S-Bench: A Hierarchical Benchmark for Streaming Spatial Intelligence in Multimodal LLMs(OVO-S-Bench:面向多模态大语言模型流式空间智能的分层基准)
[] ⚡ Qwen-Image-Flash: Beyond Objective Design(Qwen-Image-Flash:超越客观设计)
[] 🧠 M$^3$Eval: Multi-Modal Memory Evaluation through Cognitively-Grounded Video Tasks(M$^3$Eval:基于认知视频任务的多模态记忆评估)
[] 🎥 Echo-Infinity: Learning Evolving Memory for Real-Time Infinite Video Generation(回声无限:面向实时无限视频生成的可学习演化记忆)
[] 🧠 ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning(思维折叠:通过内省偏好学习折叠推理链)
[] 🧪 Benchmarks are Not Enough: RAMP for Runtime Assessing of Agentic Models in Production Systems(基准测试并不足够:用于生产系统中智能体模型运行时评估的RAMP框架)
[] ⚡ Streaming Communication in Multi-Agent Reasoning(多智能体推理中的流式通信)
[] 🎯 Self-Distilled Policy Gradient(自蒸馏策略梯度)
[] 🧠 MemTrain: Self-Supervised Context Memory Training(MemTrain:自监督上下文记忆训练)
[] 🧩 Eliciting Complex Spatial Reasoning in MLLMs through Wide-Baseline Matching(通过宽基线匹配激发多模态大语言模型中的复杂空间推理能力)
[] 🤖 MMG2Skill: Can Agents Distill In-the-Wild Guides into Self-Evolving Skills?(MMG2Skill:智能体能否从野外指南中蒸馏出自我进化的技能?)

【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
【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
