Multi-Turn Long-Form Question Answering (MT-LFQA) is a key application paradigm of Large Language Models (LLMs) in knowledge-intensive domains. However, existing benchmarks are limited to single-turn dialogue, while multi-turn dialogue benchmarks typically assess other orthogonal capabilities rather than knowledge-intensive factuality. To bridge this critical gap, we introduce **KnowMT-Bench**, a new benchmark designed to systematically evaluate MT-LFQA for LLMs across knowledge-intensive fields, including medicine, finance, and law. To faithfully assess the model’s real-world performance, KnowMT-Bench employs a dynamic evaluation setting where models generate their own multi-turn dialogue histories given logically progressive question sequences. Our experiments on a diverse suite of LLMs show a clear degradation in both factual capability and information delivery efficiency in multi-turn dialogue. We further probe the underlying causes and investigate mitigation strategies spanning domain-specific finetuning, active context denoising, and RAG. The results show domain-dependent gains from finetuning, while active context denoising and RAG more consistently mitigate these issues, with RAG notably reversing this performance degradation. These findings underscore the importance of our benchmark for evaluating LLMs’ conversational factual capabilities in real-world applications.

Reliable medical diagnosis requires a precise final prediction, which is strictly supported by observable evidence. Recent large vision-language models perform well on medical image understanding tasks, yet they often lack explicit control over which evidence to inspect before making a prediction. To facilitate flexible planning of evidence exploration, we propose MedSPA, a self-planning medical reasoning framework that separates diagnostic evidence acquisition from evidence-grounded prediction with two cooperative agents: a planning-and-reasoning agent that adaptively decides what to inspect and when to stop, and a summary agent that generates the final output from the accumulated evidence. MedSPA distills medical knowledge from strict diagnostic procedures but acquires flexibility through reinforcement learning to generate optimal outputs. Experiments on medical report generation and medical visual question answering show that MedSPA improves clinical performance, faithfulness, and controllability over existing reasoning baselines.