View of ContextFidelity-Bench: A Three-Paradigm Framework for Evaluating Multi-Turn Hallucination Behavior in Language Models | BenchCouncil Transactions on Benchmarks, Standards and Evaluations