Unverified50% confidenceFactExact time
In 2023, an Anthropic research team published a paper using sparse autoencoder technology to decompose the internal neuron activations of the Claude model into millions of interpretable conceptual features.
1
Sources
50%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAnthropic published a paper on superposition in 2022, revealing the phenomenon of individual neurons simultaneously encoding multiple concepts.73% similarUnverifiedAnthropic published the dictionary learning research Towards Monosemanticity in 2023, attempting to decompose polysemantic neurons into monosemantic features via sparse autoencoders.73% similarUnverifiedAnthropic's interpretability research team is one of the world's leading research teams, dedicated to understanding how information is represented and processed inside neural networks70% similarUnverifiedDai等人在2022年发表了论文《Knowledge Neurons in Pretrained Transformers》，通过定位和编辑特定神经元可以修改模型对特定事实的记忆68% similarUnverifiedThe field of mechanistic interpretability was largely founded by the Anthropic research team (especially Chris Olah) around 2021 through the "Transformer Circuits" series of papers.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/887187API
curl https://kongchang.com/api/v1/knowledge/claims/887187MCP
get_claim(id=887187)