hypothesis
active
hypothesis:aggregating-feature-scores-over-multiple-prompts-would-weed-out-distractor-features-and-isolate-a-cleaner-circuit-closer-to-hanna-et-alAggregating feature scores over multiple prompts would weed out distractor features and isolate a cleaner circuit closer to Hanna et al.
Limitation of single-prompt selection.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- CLT feature space is somewhat less clean than Hanna et al.'s, with some narrow special-case-lookup-like featuresassociated_withgatesDifference between sparse and neuron analyses.
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.
- Main proposal.
- Motivates nested SAE.
- Prediction for future decomposition.
- Practical implication.
- Core insight.
- We hypothesize that degraded generalization on benchmarks like MMLU may reflect the computational demands of the tasks.hypothesis0.769Connecting the paper's task-difficulty findings to prior observations of weak generalization on complex QA benchmarks.
- Discovery story.
- Scores explanations by distinguishing activating from distractor samples; weakly penalises polysemanticity.