method
active
method:logit-interpolation-stepLogit interpolation step
Samples from interpolated logits of original and abliterated models at α.
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (3)
method
- Log-linear extrapolation steprelated_toFits log compliance vs α on α with compliance <50% and extrapolates to α=1.
- The paper's method: interpolate between two models in logit space, measure compliance where common, and extrapolate to the original.
- Logit Diff AmplificationextendsPrior method extrapolating beyond the original model in logit space to surface rare behaviors.
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.
- Dot product of summed decoder vector with unembedding vector, following Hanna et al.
- Computing each feature's linear effect on output token logits via path expansion through MLP output weights and unembedding matrix
- Correlating logit weight vectors between features from different models as a measure of downstream-effect universality
- Headline result (Figure 1).
- The methodological confound identified by this paper: injection biases model toward 'YES' for any binary question regardless of content
- Training-free monitor that token-forces an eval-aware sentence and computes its probability at each CoT sentence boundary.
- Primary self-report measure: probability-weighted expected value over all ten digit-token logits, yielding a continuous rating that preserves full distributional signal
- Parameter-free loss transformation applied to each task loss to equalize scales