Dan Hendrycks @hendrycks · 21 Aug 2026
A good explanation of a model's behavior should help you make predictions in related situations. We turn this into an eval, with thousands of real behaviors found in the wild. Can interp tools help here? On average, no. 🧵 https://t.co/yN3t8G4n40
39 181Views
253Likes
32Reposts
4Replies
8Quotes
0Bookmarks
Is that a lot?
4.85×vs this author's median8 082 views is typical
80Percentile for this authorof 5 recent posts
13.5×vs 10K–100K median2 896 views is typical
87.21%Reachviews ÷ followers
0.76%Engagement rateof viewers reacted