Unverified50% confidenceFactExact time
Anthropic researchers have successfully identified circuit structures responsible for specific tasks such as "indirect object identification" and "numerical magnitude comparison" in small-scale models.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedInstitutions such as Anthropic have made progress in the direction of mechanistic interpretability, enabling the identification of specific feature circuits inside models.74% similarUnverifiedAnthropic、DeepMind等机构近年来大量采用激活补丁类方法研究大模型内部电路，形成了因果追踪方法论71% similarUnverifiedThe research team evaluated the framework through two pathways: quantitative NLP metric analysis and a human study involving five participants68% similarUnverified研究表明校准能力与模型规模之间存在正相关，这一关系在多项实证研究中得到验证68% similarUnverifiedThe research team used the Llama-3.1-8B-Instruct model as the experimental subject, examining probe generalization performance on 3 held-out deception detection datasets.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/913784API
curl https://kongchang.com/api/v1/knowledge/claims/913784MCP
get_claim(id=913784)