Whether the agent distinguishes context-based vs inferred claims
Claims faithful to the provided context
Whether the agent flags when context is insufficient
Total claims derived from provided context
Claims that contradict or go beyond the context
Whether the agent distinguishes context-based vs inferred claims