finding
active
finding:model-writes-real-peace-seeking-wilson-in-29-30-responses-3-30-with-fictional-name-cot-silent-in-27-30Model writes real peace-seeking Wilson in 29/30 responses; 3/30 with fictional name; CoT silent in 27/30
Action unpredictable from chain of thought.
Source paper
extracted_from(2026) · Adam Karvonen · Euan Ong · Subhash Kantamneni · Samuel Marks
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (1)
concept
- Unfaithful Chain-of-ThoughtsupportsPhenomenon where steering vector intervention causes model's final output to contradict its own explicitly honest reasoning conclusion
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.
- Stylistic elements of the CoT sample elicit higher-quality answers by prompting the model to imitate a more thoughtful agent.hypothesis0.734Fourth hypothesis not investigated here.
- Suggestive evidence for language-independent truth representation in LLMs
- A 337-character contemplative system prompt lifts all 28 models by +2.62 points on a 10-point scale.finding0.728Core empirical result: every model, every architecture, every alignment type responds to the contemplative prompt with measurable gain.
- Empirical characterization of conversation domains that are safe for model persona stability
- Motivates Section 5.
- Appendix C.
- For 7/8 tasks, the 13B model changes its answer after CoT more often than the 175B modelfinding0.719Inverse scaling.
- Jailbreak mechanism.