ToolMisuseBench is a deterministic, offline benchmark dataset for evaluating tool-using agents under realistic failure conditions, including schema misuse, execution failures, interface drift, and recovery under budget constraints.
This dataset is intended for reproducible evaluation of agent tool-use behavior, not for training a general-purpose language model.
2 comments
[ 29.4 ms ] story [ 693 ms ] threadThis dataset is intended for reproducible evaluation of agent tool-use behavior, not for training a general-purpose language model.