claim
active
claim:reasoning-not-being-post-hoc-does-not-guarantee-faithfulness-nor-does-being-post-hoc-exclude-faithfulnessReasoning not being post-hoc does not guarantee faithfulness, nor does being post-hoc exclude faithfulness.
Caveat on test interpretation.
Source paper
extracted_from(2023) · Tamera Lanham · Anna Chen · Ansh Radhakrishnan · Benoit Steiner +26
Neighborhood — ranked by edge-count
Papers (1)
paper
- Measuring Faithfulness in Chain-of-Thought Reasoningmentionssupports
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.
- Reasoning produced after the conclusion is already guaranteed.
- Interpretation of §3.
- Generality of methods.
- Central conclusion.
- Practical implication.
- For reasoning faithfulness, larger models may behave worse than smaller ones because they can predict answers without CoT.hypothesis0.782Motivates §3.
- Faithfulness decreases for easier problems and bigger models.
- Practical implication from Study 2 results