claim
active
claim:post-training-pushes-the-assistant-toward-lower-valence-and-lower-arousal-measured-contemplative-stancePost-training pushes the Assistant toward lower valence and lower arousal, measured contemplative stance
Interpretation of shifts.
Source paper
extracted_from(2026) · Nicholas Sofroniew · Isaac Kauvar · William Saunders · Runjin Chen +12
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (2)
finding
- Shift toward low arousal, low valence.
- Context-independent transformation.
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.
- Post-training causes the Assistant's perspective to take over increasing J-space capacity.claim0.828Base vs post-trained comparison.
- Base pretrained models show high false positive rates and achieve no net task performance on concept injection detection; post-training essential for introspection.
- Finding that base models have high false positives and no net positive performance.
- Different post-training strategies substantially influence introspection task performance; 'helpful-only' variants show higher false positives but some achieve strong net performance.
- Assertion about the role of post-training in eliciting introspection.
- Central interpretive claim and motivation for future work
- Authors' interpretive endorsement of PSM view, backed by transfer experiments
- Load-bearing summary of the paper's core finding about persona stability