ARC-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ARC-Bench experiment-stage CD:CE:RA = 25:25:50 (25 topics)
96.8Code Development Score
4
ARC-Bench 12-topic
67.5Code Development Score
3
ARC-Bench Science-Domain Tasks (B01–B07 (Biology), S01–S03 (Statistics), P01–P10 (HEP-ph))
89.8Statistics Score
3