Research question
How well designed are the benchmarks the field relies on?
Key findings
- Assessed 24 benchmarks against 46 criteria; many lack statistical reporting and documentation - Proposes a benchmark quality checklist
BetterBench
Audited widely used AI benchmarks against design and reporting best practices, finding systematic gaps.
Contribute
Spot an error, or know a record we should add? Members can propose additions and submit corrections.