🧩 Philosophy Mar 26, 2026 · daria

Test your best methods on our hard CoT interp tasks

#CoT
Less Wrong
View Channel →
Test your best methods on our hard CoT interp tasks
Source ↗ 👁 0 💬 0
Authors: Daria Ivanova, Riya Tyagi, Arthur Conmy, Neel NandaDaria and Riya are co-first authors. This work was done during Neel Nanda’s MATS 9.0. Claude helped write code and suggest edits for this post.TL;DR One of our best safety techniques right now is “just read the chain of thought”.But this isn’t always enough: can we learn more by going beyond just reading the reasoning?Yet it's such an effective technique that it's hard to tell if we have made much progress on improving methods.To help t

Comments (0)

Sign in to join the discussion

More Like This

📰
Logical Share Splitting for People who Aren't Sure if They're Intuitionists or Not
LessWrong · 2d ago
Thoughts on attending the March Against AI Extinction
LessWrong · 2d ago
Orienting to 1 year AGI timelines
LessWrong · 2d ago
📰
AGI is probably bad for startups, even if it doesn't lead to ASI
LessWrong · 2d ago
Your Research Needs More of The Void
LessWrong · 2d ago
📰
Suppose we successfully train out self-preservation
LessWrong · 2d ago