思想与社会2023 · 分享于 Hacker News
Code trained LLMs reason better, on benchmarks that have nothing to do with code
探索《Code trained LLMs reason better, on benchmarks that have nothing to do with code》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。
Ysimonw ·
Understanding GPT Tokenizers
29 赞同9 评论
Yog_kalu ·
Can Programming Languages Boost Each Other via Instruction Tuning? Yes
10 赞同1 评论
Yhhs ·
LLMs are not greater than the sum of their parts: Stanford researchers
4 赞同0 评论
Yanticrymactic ·
Ask HN: Why can GPT4 en/decode base64?
3 赞同3 评论
Ybelter ·
Chatbots Don’t Know What Stuff Isn’t
2 赞同0 评论
Yog_kalu ·
Code trained LLMs reason better, on benchmarks that have nothing to do with code
2 赞同0 评论
Ywskinner ·
AI Safety and the Age of Dislightenment
2 赞同0 评论
Ydnqthao ·
Language Models of Code Are Few-Shot Commonsense Learners
1 赞同0 评论
Ytim_sw ·
Language Models of Code Are Few-Shot Commonsense Learners
1 赞同0 评论