2026. ContextFidelity-Bench: A Three-Paradigm Framework for Evaluating Multi-Turn Hallucination Behavior in Language Models. BenchCouncil Transactions on Benchmarks, Standards and Evaluations. 6, (Sep. 2026). DOI:https://doi.org/10.66834/0azb3w20.