思想与社会2024 · 分享于 Hacker News
The Expressive Power of Transformers with Chain of Thought (Revised)
探索《The Expressive Power of Transformers with Chain of Thought (Revised)》的原文与 Hacker News 社区讨论，查看来源并加入阅读清单。
从论文，走向讨论
看看社区如何讨论这个想法。
来自 Hacker News 的原始讨论。不同的视角、疑问，以及值得继续思考的线索。
最多展示 100 条相关讨论。论文可能出现在评论中，并非原帖投稿内容。
在讨论中发现的引用
按 HN 条目顺序展示最近的 50 条引用来源。
Ycrescit_eundo ·
Something weird is happening with LLMs and chess
696 赞同474 评论
Ycolesantiago ·
Ilya Sutskever's SSI Inc raises $1B
572 赞同776 评论
Yoptimalsolver ·
30% drop in O1-preview accuracy when Putnam problems are slightly variated
560 赞同526 评论
Ychilipepperhott ·
Markov chains are the original language models
454 赞同174 评论
Yzxexz ·
Byte Latent Transformer: Patches Scale Better Than Tokens
378 赞同84 评论
Ytosh ·
Simple tasks showing reasoning breakdown in state-of-the-art LLMs
375 赞同380 评论
Ytodsacerdoti ·
The bitter lesson is coming for tokenization
303 赞同160 评论
Ykrackers ·
Chain of Thought empowers transformers to solve inherently serial problems
261 赞同184 评论
YPaulHoule ·
The wall confronting large language models
172 赞同200 评论
Ygmays ·
O1 isn't a chat model (and that's the point)
165 赞同150 评论
Ypongogogo ·
Computing Inside an AI
117 赞同68 评论
Ybelter ·
Inductive or deductive? Rethinking the fundamental reasoning abilities of LLMs
107 赞同169 评论
Ycanjobear ·
The illusion of state in state-space models
60 赞同59 评论
Yyawboakye ·
The Expressive Power of Transformers with Chain of Thought (Revised)
20 赞同1 评论
Ytosh ·
Learning more about Claude's mathematical capabilities
12 赞同2 评论
Yanonymid ·
I don't think AGI is imminent
3 赞同0 评论