Partially Verified75% confidenceReleaseExact time
In 2024, Anthropic's mechanistic interpretability team used sparse autoencoders to extract millions of interpretable concept units from the Claude model.
3
Sources
75%
Confidence
Medium-term (~90 days)
Relevance
9/5/2026
First Seen
Valid until: 12/4/2026
Sources
Related Entities
Related Claims
VerifiedAnthropic published mechanistic interpretability research on the internal feature interpretation of Claude in 2024, discovering millions of interpretable feature units.77% similarUnverifiedIn 2024, Anthropic used sparse autoencoders to extract 34 million features from Claude Sonnet and located the internal circuits corresponding to specific concepts.73% similarUnverifiedIn 2024, Anthropic successfully decomposed millions of interpretable features in the Claude model using sparse autoencoders (SAE).71% similarUnverifiedIn 2024, Anthropic extended mechanistic interpretability methods to Claude 3 Sonnet-level models, identifying millions of interpretable features, including features corresponding to concepts such as deception and sycophancy.69% similarUnverifiedIn a paper published by Anthropic in 2023, sparse autoencoders were used to find tens of thousands of interpretable features in Claude, including emotional states and programming language types.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/859093API
curl https://kongchang.com/api/v1/knowledge/claims/859093MCP
get_claim(id=859093)