Benchmark RadarAI BENCHMARK PROFILE
OpenSciToolBench
OpenSciToolBench is a benchmark with 900 tasks across four difficulty levels for evaluating LLM agents in open-world scientific tool acquisition.
- Released
- 2026-07-30
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
The benchmark supports a specific agent system's evaluation and lacks a standalone comparison path.
Motivation
Large language model (LLM) agents have been increasingly adopted in scientific research for organizing and invoking specialized computational tools.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.