主播
节目简介
来源:小宇宙
你有没有想过,AI也能学会“未卜先知”,通过预判下一秒的需求,来打破效率瓶颈吗?本期节目,我们将从几篇最新的AI论文出发,看一看科学家们是如何教会AI这些神奇的“超能力”的。我们会聊到,如何通过一位“过程教练”的精妙指导,让AI学会分析自己的“草稿纸”,而不是只看最终答案。我们还会探讨,如何给AI配上一副“全局透视镜”,让它在处理复杂任务时不再有“上下文盲区”。更神奇的是,我们还会发现,一个起点不高的小模型,是怎样在顶级教练的指导下,成长为能写长篇小说的“耐力型选手”的。最后,我们还会从“三个和尚没水喝”的故事,聊到上万台计算机如何通过“慢一步”的智慧,实现高效协同。准备好了吗?让我们一起探索AI思考方式的进化!
00:00:54 AI的远见,如何让模型学会“未卜先知”?
00:05:50 给AI请个好教练,小个子也能跑马拉松
00:11:19 人多,如何才能力量大?
00:16:34 AI也需要一位“过程教练”
00:22:12 你的说明书,AI都看不出漏洞?
本期介绍的几篇论文:
[CL] SparDA: Sparse Decoupled Attention for Efficient Long-Context LLM Inference
[NVIDIA & Thinking Machines Lab & ByteDance Seed]
https://arxiv.org/abs/2606.04511
---
[CL] POLARIS: Guiding Small Models to Write Long Stories
[University of Maryland & Google DeepMind]
https://arxiv.org/abs/2606.04095
---
[LG] Near-Optimal Decentralized Stochastic Convex Optimization over Networks
[Tel Aviv University]
https://arxiv.org/abs/2606.04757
---
[CL] Read the Trace, Steer the Path: Trajectory-Aware Reinforcement Learning for Diffusion Language Models
[Microsoft AI]
https://arxiv.org/abs/2606.04396
---
[LG] Context-as-a-Service: Surfacing Cross-File Dependency Chains for LLM-Generated Developer Documentation
[Meta]
https://arxiv.org/abs/2606.04397
00:00:54 AI的远见,如何让模型学会“未卜先知”?
00:05:50 给AI请个好教练,小个子也能跑马拉松
00:11:19 人多,如何才能力量大?
00:16:34 AI也需要一位“过程教练”
00:22:12 你的说明书,AI都看不出漏洞?
本期介绍的几篇论文:
[CL] SparDA: Sparse Decoupled Attention for Efficient Long-Context LLM Inference
[NVIDIA & Thinking Machines Lab & ByteDance Seed]
https://arxiv.org/abs/2606.04511
---
[CL] POLARIS: Guiding Small Models to Write Long Stories
[University of Maryland & Google DeepMind]
https://arxiv.org/abs/2606.04095
---
[LG] Near-Optimal Decentralized Stochastic Convex Optimization over Networks
[Tel Aviv University]
https://arxiv.org/abs/2606.04757
---
[CL] Read the Trace, Steer the Path: Trajectory-Aware Reinforcement Learning for Diffusion Language Models
[Microsoft AI]
https://arxiv.org/abs/2606.04396
---
[LG] Context-as-a-Service: Surfacing Cross-File Dependency Chains for LLM-Generated Developer Documentation
[Meta]
https://arxiv.org/abs/2606.04397