TroubleshootingBench

TroubleshootingBench is an OpenAI internal short-answer benchmark built from expert-written wet-lab procedures and realistic execution errors to test tacit biological troubleshooting knowledge.

id troubleshootingbench · max 1 · 1 models reported