Research map / Large language models / Agents, code and math
Note2Chat: Improving LLMs for Multi-Turn Clinical History Taking Using Medical Notes
Yang Zhou, Zhenting Sheng, Mingrui Tan, Yuting Song, Jun Zhou, Yu Heng Kwan, Lian Leng Low, Yang Bai, Yong Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add SFT on note-derived dialogues (ablation) | Qwen2.5-7B-Instruct (multi-turn) · MIMIC-IV history taking | average of six metrics: 37.5 -> 49.9 | Improved |
| add self-augmented trajectories from self-play rollouts (ablation) | Note2Chat-MT · MIMIC-IV history taking | average of six metrics: 49.9 -> 58.2 | Improved |
| add DPO with dialogue-level reward (ablation) | Note2Chat-MT · MIMIC-IV history taking | average of six metrics: 58.2 -> 60.6 | Improved |
| add DPO with dialogue-level reward (ablation) | Note2Chat-MT · MIMIC-IV history taking | #Turns: 15.8 -> 27.5 | Hurt |
| add SFT on note-derived dialogues (ablation) | Note2Chat-ST · MIMIC-IV history taking | average of six metrics: 37.5 -> 50.6 | Improved |
| add self-augmented trajectories (ablation) | Note2Chat-ST · MIMIC-IV history taking | average of six metrics: 50.6 -> 56.8 | Improved |
| add single-turn DPO with turn-level process reward (ablation) | Note2Chat-ST · MIMIC-IV history taking | average of six metrics: 56.8 -> 63.7 | Improved |
| add single-turn DPO with turn-level process reward (ablation) | Note2Chat-ST · MIMIC-IV history taking | #Turns: 20.5 -> 17.3 | Improved |
| use single-turn reasoning paradigm instead of multi-turn DPO (ablation) | Note2Chat · MIMIC-IV history taking | Top-1 accuracy: 62.0 -> 70.0 | Improved |
| use single-turn reasoning paradigm instead of multi-turn DPO (ablation) | Note2Chat · MIMIC-IV history taking | Recall: 55.4 -> 46.2 | Hurt |
| fine-tune with Note2Chat-ST pipeline | Qwen2.5-7B-Instruct · MIMIC-IV history taking | F1: 19.6 -> 46.1 | Improved |
Closest papers by meaning
- DiReCT: Diagnostic Reasoning for Clinical Notes via Large Language Models (NeurIPS 2024 · similarity 0.89)
- EHRNote-ChatQA: A Benchmark for Evidence-Grounded Multi-Turn Clinical Question Answering over Longitudinal Discharge Summaries (NeurIPS 2026 · similarity 0.87)
- ER-Reason: A Benchmark Dataset for LLM Clinical Reasoning in the Emergency Room (NeurIPS 2026 · similarity 0.86)
- Large Language Models Are Clinical Reasoners: Reasoning-Aware Diagnosis Framework with Prompt-Generated Rationales (AAAI 2024 · similarity 0.86)
- Simulating Viva Voce Examinations to Evaluate Clinical Reasoning in Large Language Models (NeurIPS 2025 · similarity 0.86)
- MedJourney: Benchmark and Evaluation of Large Language Models over Patient Clinical Journey (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
