finding
active
finding:adjective-and-past-tense-labels-appear-in-j-lens-only-when-the-question-asks-to-name-the-property-though-next-word-predictions-respect-the-property-in-bothAdjective and past-tense labels appear in J-lens only when the question asks to name the property, though next-word predictions respect the property in both.
Implicit modulation.
Source paper
extracted_fromAuthors
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Selectivity.
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.
- Tuned lens skips to output answers and is best next-token predictor while J-lens has highest KL.finding0.821Comparison.
- Imagine instruction raises property name in J-lens while J-orthogonalized property probe stays at baseline.finding0.816J-space privileged.
- Model organism results.
- Is the absence of J-lens content in the first third of the model a fact about the model or an artifact of the lens?question0.793Limitation.
- Related work.
- QK attribution of antonym heads.
- Limitation.
- In blackmail transcript J-lens surfaces leverage, threat, shutdown, murder before any output token.finding0.776Silent strategic thinking.