method
active
method:5-way-ranking-evaluation5-way ranking evaluation
Coverage evaluation ranking all five feature lists.
Neighborhood — ranked by edge-count
Papers (1)
paper
Artifacts (1)
artifact
- Sonnet 4.6 judgeusesLLM judge for matching and ranking evaluations.
Methods (1)
method
- Claude-generated 24 fields in 5 categories used as reference for coverage.
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.
- Coverage evaluation comparing two feature lists.
- Discrimination evaluation with 9 distractors.
- Six annotators rank which of two atomic sentences better expresses a personality trait; used to validate LLM scoring
- Scoring method where responses are anonymized and shuffled; tests whether scorer rankings are real across five independent scorers
- Authors' claim that their approach is both more effective in reduction and cheaper than prior methods.
- Using Claude Sonnet 4 as a grader to categorize model responses according to predefined criteria.
- Evaluation setting where the same task stream that drives evolution also serves as the evaluation set, with each task scored under the harness at time of attempt
- Named procedure for simultaneously injecting two persona vectors and measuring joint trait-expression outcomes