Unverified70% confidenceFactExact time
Anthropic 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.
2
Sources
70%
Confidence
Long-term
Relevance
7/19/2026
First Seen
Sources
Related Claims
UnverifiedIn 2024, Anthropic successfully decomposed millions of interpretable features in the Claude model using sparse autoencoders (SAE).78% similarUnverifiedAnthropic于2023年发表论文《Towards Monosemanticity》，首次在百万参数级别模型中识别出数万个单语义特征74% similarVerifiedAnthropic published mechanistic interpretability research on the internal feature interpretation of Claude in 2024, discovering millions of interpretable feature units.74% similarUnverifiedThe State of AI Report 2026 shows that Claude guided about 26% of the research directions in Anthropic's core model R&D, while in February 2026 this figure was still less than 1%.67% 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66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/561825API
curl https://kongchang.com/api/v1/knowledge/claims/561825MCP
get_claim(id=561825)