claim
active
claim:the-model-mixes-a-mechanism-tagging-petit-as-the-word-to-invert-with-one-searching-for-any-plausible-adjectiveThe model mixes a mechanism tagging 'petit' as the word to invert with one searching for any plausible adjective
Antonym interpretation.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Findings (2)
finding
- QK attribution of antonym heads.
- Steering validation for antonym mechanism.
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.
- Evidence that core representations like preferences are persona-relative, supporting claim that personas gate content of representations
- Antonym example.
- All models exhibit above-baseline representation of the think word when instructed to think about itfinding0.738In the intentional control experiment, all tested models show above-zero cosine similarity to the think word's concept vector.
- The model is able to express many subforms of a persona, with different elicitation methods eliciting different manifestationshypothesis0.736PSM-derived hypothesis supported by discourse-type facet analysis
- Suggestive evidence for language-independent truth representation in LLMs
- Articulates why a one-layer transformer with MLP is the appropriate starting target for mechanistic interpretability
- New view introduced by this paper: a mind is the union of all instance-persona segments that activate the same persona region of a given model
- Second of two central questions motivating the paper