Terminal-Bench-Science 0.1: evaluating AI agents in scientific research workflows
Terminal-Bench-Science evaluates AI agents on workflows from researchers' own work. Scientists, not model developers or data vendors, set the bar for scientific capability in AI.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open original at terminal-bench-science.ai