question
active
question:can-latent-codes-from-saes-or-temporal-feature-analyzers-capture-the-in-context-graph-phenomenonCan latent codes from SAEs or Temporal Feature Analyzers capture the in-context graph phenomenon?
§6.3 question.
Source paper
extracted_from(2025) · Lubana, Ekdeep Singh · Rager, Can · Hindupur, Sai Sumedh R. · Costa, Valerie +12
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Fig. 9b.
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.
- §7.2 assertion.
- Out-of-distribution generalization of SAE features.
- Cited as enabling precise behavioral control through SAE features, extending the same methodological line
- Extension of mechanistic interpretability findings to the metacognitive domain
- Proposition 4.1 central claim.
- Downside of abstraction.
- Motivating question.