
时长:
8分钟
播放:
88
发布:
2周前
主播...
简介...
本期的 15 篇论文如下:
[00:23] 🚀 AgentFly: Fine-tuning LLM Agents without Fine-tuning LLMs(AgentFly:无需微调LLM即可微调LLM智能体)
[00:48] 🐕 ODYSSEY: Open-World Quadrupeds Exploration and Manipulation for Long-Horizon Tasks(ODYSSEY:开放世界四足机器人长周期任务探索与操作)
[01:24] 📈 Beyond Pass@1: Self-Play with Variational Problem Synthesis Sustains RLVR(超越Pass@1:变分问题合成的自博弈策略持续提升RLVR)
[01:51] 🗑 CRISP: Persistent Concept Unlearning via Sparse Autoencoders(CRISP:基于稀疏自编码器的持久概念消除)
[02:21] 🔍 Selective Contrastive Learning for Weakly Supervised Affordance Grounding(选择性对比学习用于弱监督动作功能区域定位)
[02:49] 🏆 AetherCode: Evaluating LLMs' Ability to Win In Premier Programming Competitions(AetherCode:评估LLM在顶级编程竞赛中的获胜能力)
[03:19] 👁 EgoTwin: Dreaming Body and View in First Person(EgoTwin:第一人称视角的身体与视野生成)
[03:46] 🤔 Do What? Teaching Vision-Language-Action Models to Reject the Impossible(做什么?教导视觉-语言-动作模型拒绝不可能)
[04:14] 🩺 End-to-End Agentic RAG System Training for Traceable Diagnostic Reasoning(端到端智能体RAG系统训练,实现可追溯的诊断推理)
[04:40] ⚡ TPLA: Tensor Parallel Latent Attention for Efficient Disaggregated Prefill \& Decode Inference(TPLA:用于高效解耦预填充与解码推理的张量并行潜在注意力)
[05:06] 🤖 AgentScope 1.0: A Developer-Centric Framework for Building Agentic Applications(AgentScope 1.0:一个以开发者为中心的智能体应用构建框架)
[05:37] 🔄 RotaTouille: Rotation Equivariant Deep Learning for Contours(RotaTouille:轮廓的旋转等变深度学习)
[06:04] 🤔 InMind: Evaluating LLMs in Capturing and Applying Individual Human Reasoning Styles(InMind:评估LLM在捕获和应用个体人类推理风格方面的能力)
[06:28] 🚀 CARFT: Boosting LLM Reasoning via Contrastive Learning with Annotated Chain-of-Thought-based Reinforced Fine-Tuning(CARFT:通过结合带标注思维链的强化微调与对比学习提升大型语言模型推理能力)
[06:54] ✏ Sketch3DVE: Sketch-based 3D-Aware Scene Video Editing(Sketch3DVE:基于草图的3D感知场景视频编辑)
【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
[00:23] 🚀 AgentFly: Fine-tuning LLM Agents without Fine-tuning LLMs(AgentFly:无需微调LLM即可微调LLM智能体)
[00:48] 🐕 ODYSSEY: Open-World Quadrupeds Exploration and Manipulation for Long-Horizon Tasks(ODYSSEY:开放世界四足机器人长周期任务探索与操作)
[01:24] 📈 Beyond Pass@1: Self-Play with Variational Problem Synthesis Sustains RLVR(超越Pass@1:变分问题合成的自博弈策略持续提升RLVR)
[01:51] 🗑 CRISP: Persistent Concept Unlearning via Sparse Autoencoders(CRISP:基于稀疏自编码器的持久概念消除)
[02:21] 🔍 Selective Contrastive Learning for Weakly Supervised Affordance Grounding(选择性对比学习用于弱监督动作功能区域定位)
[02:49] 🏆 AetherCode: Evaluating LLMs' Ability to Win In Premier Programming Competitions(AetherCode:评估LLM在顶级编程竞赛中的获胜能力)
[03:19] 👁 EgoTwin: Dreaming Body and View in First Person(EgoTwin:第一人称视角的身体与视野生成)
[03:46] 🤔 Do What? Teaching Vision-Language-Action Models to Reject the Impossible(做什么?教导视觉-语言-动作模型拒绝不可能)
[04:14] 🩺 End-to-End Agentic RAG System Training for Traceable Diagnostic Reasoning(端到端智能体RAG系统训练,实现可追溯的诊断推理)
[04:40] ⚡ TPLA: Tensor Parallel Latent Attention for Efficient Disaggregated Prefill \& Decode Inference(TPLA:用于高效解耦预填充与解码推理的张量并行潜在注意力)
[05:06] 🤖 AgentScope 1.0: A Developer-Centric Framework for Building Agentic Applications(AgentScope 1.0:一个以开发者为中心的智能体应用构建框架)
[05:37] 🔄 RotaTouille: Rotation Equivariant Deep Learning for Contours(RotaTouille:轮廓的旋转等变深度学习)
[06:04] 🤔 InMind: Evaluating LLMs in Capturing and Applying Individual Human Reasoning Styles(InMind:评估LLM在捕获和应用个体人类推理风格方面的能力)
[06:28] 🚀 CARFT: Boosting LLM Reasoning via Contrastive Learning with Annotated Chain-of-Thought-based Reinforced Fine-Tuning(CARFT:通过结合带标注思维链的强化微调与对比学习提升大型语言模型推理能力)
[06:54] ✏ Sketch3DVE: Sketch-based 3D-Aware Scene Video Editing(Sketch3DVE:基于草图的3D感知场景视频编辑)
【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
评价...
空空如也
小宇宙热门评论...
暂无小宇宙热门评论