Back to directory
ScienceAgentBench
Free140[ICLR'25] ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery
About
ScienceAgentBench evaluates how well AI language agents perform scientific research tasks using real data. It's built for researchers testing AI systems in discovery workflows. The benchmark includes standardized metrics that measure whether agents actually produce valid scientific results, not just plausible-sounding outputs.
Key Features
- ai4science
- bioinformatics
- chemoinformatics
- code-generation
- cognitive-neuroscience
Pricing
FreeOpen source. You supply your own LLM API keys.
Categories
ResearchData Analysis