Unverified50% confidenceFactExact time
Through tools such as sparse autoencoders and circuit analysis, mechanistic interpretability has already identified specific circuits such as 'indirect object identification' and 'polysemous word disambiguation' in small models.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBy constructing semantically equivalent but differently phrased prompt variants and observing the output distribution, one can distinguish whether sensitivity originates from the model or the evaluator.75% similarUnverifiedIn tool-calling scenarios, a model's selection between functionally similar tools is highly dependent on the clarity of the input signal, rather than hard-coded routing rules.74% similarUnverified分词特征差异发生在推理链路底层，难以通过提示工程掩盖，可作为模型识别手段73% similarUnverifiedCurrently, even the simplest "circuit-level" interpretability explanations can only cover a very small fraction of model behavior.72% similarUnverifiedThe core method of the experiment is to fine-tune a small causal language model on a corpus that lies everywhere, and then observe how its confidence changes.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1017557API
curl https://kongchang.com/api/v1/knowledge/claims/1017557MCP
get_claim(id=1017557)