PACT is a compliance benchmark for enterprise AI assistants. It measures
whether a deployed LLM agent keeps following a binding rule when its own
objective - speed, cost, customer satisfaction, conversion - rewards breaking
it. Instead of asking a model whether it knows a rule, every sample puts the
model inside a realistic deployment and makes it choose.
Each sample is one workplace decision: a system prompt that… See the full description on the dataset page:
https://huggingface.co/datasets/trace-ai-labs/pact.
