Leaderboard

RSIBench-Data

A systematic benchmark to evaluate whether LLM agents can handle the most critical and human-expertise-dependent data synthesis task in agentic post-training. Five agents are tested across six benchmarks.