concept
active
concept:copy-suppression-anti-inductionCopy suppression / anti-induction
Behavior of IOI inhibition head and GPT-2 L10H7.
Neighborhood — ranked by edge-count
Papers (1)
paper
- HeadVis: An Interactive Tool For Investigating Attention Headsintroducesmentions
Frameworks (1)
framework
- IOI circuitassociated_withPrior circuit containing an inhibition head.
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.
- Neural phenomenon reproduced by active inference model: reduced response to repeated stimuli.
- Head suppresses attention between related items, e.g., country and its own cities.
- Maintenance of tissue-level order; disruption of bioelectric networks can lead to tumorigenesis, which can be reversed by electrical normalization.
- Induction performed in the feature basis rather than token space.
- Emerging framework that seeks invariants between evolution and learning; cited as future direction.
- The phenomenon where activating an emotion feature leads to subsequent below-baseline activation of that feature
- Actions taken by the model to undermine the AI developer, such as weight exfiltration, lying to contractors, or helping whistleblowers
- Central interpretation of induction case study.