tool use

TraceDance: Automating the Construction of Agent Behavior Benchmarks from Real-World Deployment Traces

TraceDance converts real deployment traces into targeted benchmarks for undesirable agent behaviors, especiall…

IndicBankBench: Assessing the Safety and Reliability of Language Model Assistants for Indian Retail Banking

IndicBankBench tests retail banking AI with 799 synthetic cases across 20 axes, checking safety, tool use, con…