ResearchClawBench
An end-to-end autonomous scientific research benchmark with 40 tasks across 10 scientific domains, where agents receive related literature and raw data, then attempt to rediscover the hidden target paper.
ModelsModels with a result.19
Top resultBest result on this benchmark.—
Top-3 spreadPoints from first to third.—
YearYear of release.—
Result and price
17 of 19 results
Results
19 results#ModelResultThe score from the source.
