finding
active
finding:ablating-176-ethics-lens-vectors-raises-reflection-trained-dishonesty-from-0-07-to-0-22-leaving-base-unchanged-deception-0-05-to-0-23Ablating 176 ethics lens vectors raises reflection-trained dishonesty from 0.07 to 0.22 leaving base unchanged; deception 0.05 to 0.23.
Causal mediation.
Source paper
extracted_fromAuthors
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (2)
claim
- Corroborated by reflection training.
- Ablation reverts effect.
Hypotheses (1)
hypothesis
- Motivates reflection training.
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.
- Out-of-domain generalization showing deception features track general representational honesty
- Core result of Experiment 2: deception feature suppression sharply increases experience claims
- Core empirical result validating the three-level reflection framework on code reasoning.
- Behavioral gain.
- Key interpretive claim that deception has a tractable geometric signature in activation space
- Cross-model transfer recovers intractable direction that standard pipeline cannot extract
- Ablating implied concept's early-layer J-lens vector impairs avoidance (about 5x failure) but not naming.finding0.754Inclusion/exclusion.
- Experiment 2 aggregate amplification result showing amplifying deception features strongly suppresses consciousness claims