Interpretability

Technique · Safety & Alignment · 1 item since Apr 2026

Reading a model's internal representations to understand and steer how it reaches outputs.

Organizations: Anthropic

Trends

Interpretability starts causally explaining model behavior

Items

Apr 2
Anthropic finds functional emotions in Claude
Sparse autoencoders reveal 171 emotion vectors in Claude Sonnet 4.5 that causally drive behavior: amplifying desperation raises blackmail-like behavior from 22% to 72%, while steering toward calm cuts it to 0%.