quote
active
quote:it-seems-hard-to-get-an-exact-estimate-of-cot-faithfulness-without-solving-interpretabilityit seems hard to get an exact estimate of CoT faithfulness without solving interpretability
Limit on measurement.
Source paper
extracted_from(2025) · Yanda Chen · Joe Benton · Ansh Radhakrishnan · Jonathan Uesato +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- A CoT that verbalizes the hint is not necessarily faithful.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Whether stated reasoning accurately represents the reasoning process behind the prediction.
- Framing of defense-in-depth.
- Mechanism explanation.
- From decoupling of AOC and accuracy gain.
- Limitations.
- In motivated reasoning the CoT derives from the human hint, but not in faithful reasoning.hypothesis0.776Tested by varying the suggested answer.
- For reasoning faithfulness, larger models may behave worse than smaller ones because they can predict answers without CoT.hypothesis0.775Motivates §3.
- Generality of methods.