Research map / Large language models / Agents, code and math
Beyond N-grams: A Hierarchical Reward Learning Framework for Clinically-Aware Medical Report Generation
Yuan Wang, Shujian Gao, Jiaxiang Liu, Songtao Jiang, Xia Haoxiang, Xiaotian Zhang, Zhaolu Kang, Yemin Wang, Zuozhu 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 semantic-level LLM-verifier reward on top of token and concept rewards (ablation) | HiMed-3B · MIMIC-CXR | ROUGE-L: 0.258 -> 0.271 | Improved |
| add semantic-level LLM-verifier reward on top of token and concept rewards (ablation) | HiMed-3B · CheXpert | ROUGE-L: 0.138 -> 0.157 | Improved |
| add semantic-level LLM-verifier reward on top of token and concept rewards (ablation) | HiMed-3B · IU-Xray | ROUGE-L: 0.301 -> 0.319 | Improved |
| use dynamic linear reward weight adjustment instead of fixed weights (ablation) | HiMed-3B · MIMIC-CXR | RaTEScore: 0.525 -> 0.544 | Improved |
| set alpha_min 0.1 rather than 0.3 in linear schedule (ablation) | HiMed-3B · MIMIC-CXR | RaTEScore: 0.535 -> 0.544 | Improved |
| set alpha_min 0.1 rather than 0.0 in linear schedule (ablation) | HiMed-3B · MIMIC-CXR | RaTEScore: 0.541 -> 0.544 | No clear effect |
| use larger Qwen3-30B-A3B verifier instead of 7-8B verifiers (ablation) | HiMed-3B · CheXpert | ROUGE-L: 0.135 -> 0.157 | Improved |
| use small 7-8B verifiers instead of no semantic reward (ablation) | HiMed-3B · CheXpert | ROUGE-L: 0.138 -> 0.134 | No clear effect |
| add RL fine-tuning after cold-start SFT (ablation) | HiMed-3B · MIMIC-CXR, CheXpert, IU-Xray | report quality | Improved |
| HiMed-3B vs strongest baseline out of domain | HiMed-3B · PadChest-GR (out-of-domain, English) | RaTEScore: 0.348 -> 0.371 | Improved |
Closest papers by meaning
- Large Language Models Are Clinical Reasoners: Reasoning-Aware Diagnosis Framework with Prompt-Generated Rationales (AAAI 2024 · similarity 0.86)
- Memorize and Rank: Elevating Large Language Models for Clinical Diagnosis Prediction (AAAI 2025 · similarity 0.86)
- PromptMRG: Diagnosis-Driven Prompts for Medical Report Generation (AAAI 2024 · similarity 0.85)
- CURE: Curriculum-guided Multi-task Training for Reliable Anatomy Grounded Report Generation (CVPR 2026 · similarity 0.85)
- MEPNet: Medical Entity-Balanced Prompting Network for Brain CT Report Generation (AAAI 2025 · similarity 0.84)
- ReFINE: A Reward-Based Framework for Interpretable and Nuanced Evaluation of Radiology Report Generation (AAAI 2026 · similarity 0.84)
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.
