Unverified60% confidenceFactExact time
In 2024, Anthropic successfully decomposed millions of interpretable features in the Claude model using sparse autoencoders (SAE).
2
Sources
60%
Confidence
Long-term
Relevance
8/20/2026
First Seen
Sources
Related Claims
UnverifiedAnthropic published "Towards Monosemanticity" in 2023 and "Scaling and evaluating sparse autoencoders" in 2024, with researchers using SAEs to identify millions of interpretable features in models such as Claude.78% similarUnverifiedAnthropic在2023年发表的研究表明，通过SAE可以从Claude模型中提取出对应'金门大桥'、'代码错误'等具体概念的单元特征78% similarUnverifiedBetween 2023 and 2024, Anthropic identified millions of interpretable features in the Claude model, including the Golden Gate Bridge feature experiment77% similarVerifiedAnthropic published mechanistic interpretability research on the internal feature interpretation of Claude in 2024, discovering millions of interpretable feature units.76% similarUnverifiedAnthropic introduced tool_use capability for the Claude model in 2024.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/780102API
curl https://kongchang.com/api/v1/knowledge/claims/780102MCP
get_claim(id=780102)