claim
active
claim:averaged-across-examples-memorization-directions-are-flatter-than-generalizing-directions-inverting-the-per-example-view-that-memorized-points-are-sharpAveraged across examples, memorization directions are flatter than generalizing directions, inverting the per-example view that memorized points are sharp
Core novel interpretation.
Source paper
extracted_from(2025) · Merullo, Jack · Vatsavaya, Srihita · Bushnaq, Lucius · Lewis, Owen
Neighborhood — ranked by edge-count
Papers (1)
paper
- From Memorization to Reasoning in the Spectrum of Loss Curvaturementionssupports
Claims (1)
claim
- Explains population vs instance curvature.
Questions (1)
question
- Aggregate analysis motivation.
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.
- Consistent with prior work.
- Theoretical limitation identified by the authors distinguishing reflection from stylistic tasks.
- Features not merely relayed via residual connections.
- Abstract thesis.
- Key methodological claim: MM probes are both competitive in accuracy and superior in causal influence
- Section 6.
- Motivating hypothesis for Section 5's investigation of prompt template effects.
- Everything is vague to a degree you do not realize till you have tried to make it precise.concept0.756Epigraph motivating the necessity of precise semantic specification in programming