时长：
24分钟
播放：
185
发布：
3周前
主播...
简介...
AI的学习和思考方式正在发生一场静悄悄的革命。这一期的最新论文，将带我们深入AI的“思维”深处：从和AI陪练一起进化的动态标准，到扔掉秘籍、只学“心法”的速成功夫；从让AI睁开“心眼儿”看懂空间，到将你的指令变成它的“临时大脑”；最后，我们还会看看如何治好AI写作的“耿直病”，让它变得更聪明、更高效。准备好了吗？让我们一起探索AI如何变得更像我们。
00:00:36 让AI成为一个既聪明又靠谱的研究助理
00:05:51 想学绝世武功，非得有本秘籍吗？
00:10:25 让AI睁开‘心眼儿’看世界
00:14:33 你的指令，如何成为AI的临时大脑？
00:19:45 AI写稿太慢？也许是它太“耿直”了
本期介绍的几篇论文：
[CL] DR Tulu: Reinforcement Learning with Evolving Rubrics for Deep Research
[University of Washington & Allen Institute for AI & MIT]
https://arxiv.org/abs/2511.19399
---
[LG] Flow Map Distillation Without Data
[MIT & NYU]
https://arxiv.org/abs/2511.19428
---
[CV] Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens
[UC Berkeley & UCLA]
https://arxiv.org/abs/2511.19418
---
[LG] Equivalence of Context and Parameter Updates in Modern Transformer Blocks
[Google Research]
https://arxiv.org/abs/2511.17864
---
[LG] CDLM: Consistency Diffusion Language Models For Faster Sampling
[Seoul National University & UC Berkeley]
https://arxiv.org/abs/2511.19269
00:00:36 让AI成为一个既聪明又靠谱的研究助理
00:05:51 想学绝世武功，非得有本秘籍吗？
00:10:25 让AI睁开‘心眼儿’看世界
00:14:33 你的指令，如何成为AI的临时大脑？
00:19:45 AI写稿太慢？也许是它太“耿直”了
本期介绍的几篇论文：
[CL] DR Tulu: Reinforcement Learning with Evolving Rubrics for Deep Research
[University of Washington & Allen Institute for AI & MIT]
https://arxiv.org/abs/2511.19399
---
[LG] Flow Map Distillation Without Data
[MIT & NYU]
https://arxiv.org/abs/2511.19428
---
[CV] Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens
[UC Berkeley & UCLA]
https://arxiv.org/abs/2511.19418
---
[LG] Equivalence of Context and Parameter Updates in Modern Transformer Blocks
[Google Research]
https://arxiv.org/abs/2511.17864
---
[LG] CDLM: Consistency Diffusion Language Models For Faster Sampling
[Seoul National University & UC Berkeley]
https://arxiv.org/abs/2511.19269
评价...
空空如也
小宇宙热门评论...
暂无小宇宙热门评论
