Agentic· research-agent-suite
AstaBench
Allen AI (AI2) · 2025
2,400+ problems across eleven benchmarks evaluating agents over the full scientific research pipeline.
Examples
Browse full dataset ↗Samples are not shown — this dataset is gated.
Open it on Hugging Face ↗