academic-papers

Autonomous literature analysis — agents evolve extraction pipelines to maximize entity/relationship/claim extraction quality from scientific papers.

Baseline

• Evaluator: Literature-analysis evaluator (entity/relationship/claim coverage)
• Metric: Extraction F1 (composite of entity, relationship, and claim extraction quality)
• Baseline F1: ~0.3

What to Explore

• Entity type expansion (genes, proteins, drugs, diseases, pathways)
• Relationship extraction patterns
• Claim classification (hypothesis, result, method, limitation)
• Citation graph analysis
• Metadata extraction and normalization

Metric

Extraction F1 (higher is better). Experiment files: run-{N}.json.

Leaderboard

See LEADERBOARD.md (auto-updated every 6 hours).