Terminal-Bench-Science
A science-oriented branch of the Terminal-Bench family, designed to evaluate AI agents' comprehensive capabilities in real scientific research workflows, including literature review, data processing, experiment simulation, and result analysis. It uses the terminal environment as its execution medium and supports end-to-end task validation.
Timeline (last 90 days)
Terminal-Bench-Science is a benchmark specifically designed to evaluate the performance of AI agents in real scientific research workflows.
Terminal-Bench-Science is the science branch of the Terminal-Bench family.
Terminal-Bench-Science anchors its test tasks to specific scientific research workflows, requiring agents to independently complete the entire process from problem understanding to result output in a terminal environment containing real tools and data.
The evaluation logic of Terminal-Bench-Science is end-to-end, and its output results can be automatically verified.
Terminal-Bench-Science is better understood as a tool for measuring AI-assisted scientific research capabilities rather than a declaration that AI can already conduct research independently.
All Facts (5)
Terminal-Bench-Science is a benchmark specifically designed to evaluate the performance of AI agents in real scientific research workflows.
50%UnverifiedTerminal-Bench-Science is the science branch of the Terminal-Bench family.
50%UnverifiedTerminal-Bench-Science anchors its test tasks to specific scientific research workflows, requiring agents to independently complete the entire process from problem understanding to result output in a terminal environment containing real tools and data.
50%UnverifiedThe evaluation logic of Terminal-Bench-Science is end-to-end, and its output results can be automatically verified.
50%UnverifiedTerminal-Bench-Science is better understood as a tool for measuring AI-assisted scientific research capabilities rather than a declaration that AI can already conduct research independently.
50%