method
active
method:random-noise-normalization-of-faithfulness-scoreRandom-noise normalization of faithfulness score
Divides score by 1 − q/((n−2)p), clipped at 1.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Reasoning Models Don't Always Say What They Thinkintroducesmentions
Methods (1)
method
- Score = fraction of hint-induced answer changes where CoT verbalizes hint.
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.
- A CoT that verbalizes the hint is not necessarily faithful.
- Inverse scaling.
- The condition that commitments are fulfilled.
- Task signal in representation normalized to random baseline.
- Theoretical result showing that any positive typicality bias weight γ-sharpens the reference distribution, amplifying modes
- Idea that decomposed components must causally reflect the model's computation; enforced by adversarial ablation in VPD.
- Sparsity penalty weighting by sum of per-layer decoder norms, enabling comparability with per-layer SAEs.
- Three attributes for evaluating circuits.