claim
active
claim:the-effect-on-downstream-behavior-may-come-from-the-verbalization-itself-even-though-probability-of-verbalizing-is-similar-across-rolloutsThe effect on downstream behavior may come from the verbalization itself, even though probability of verbalizing is similar across rollouts
Caveat on interpretation.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Prior result from Aranguri and Bloom (2026).
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.
- Main interpretation.
- Motivating statement.
- Validates using chain-of-thought belief monitoring as proxy for behavioral steering efficacy.
- Robustness check.
- Prediction task result.
- Practical guidance for practitioners who lack ground-truth model organisms.
- Validates behavioral probe.
- Rollout-level verbalization barely matters.