Research map / Vision-language and multimodal / Multimodal LLMs
Linguistically-Aligned and Visually-Grounded Preference Optimization for Clinically-Augmented Medical Report Generation
Qiang Hu, Yuxuan Luo, Yingjie Guo, Hao Wang, Qimei Wang, Qiang Li, Zhiwei Wang
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 |
|---|---|---|---|
| mitigate explicit errors without ECD module (ablation) | RADAR · MIMIC-CXR (RADAR) | 14Ma-F1: 0.460 -> 0.452 | Hurt |
| add ECD module to explicit-error mitigation (ablation) | RADAR + MEE · MIMIC-CXR (RADAR) | 14Ma-F1: 0.452 -> 0.482 | Improved |
| add knowledge-graph verification to latent-risk mitigation (ablation) | RADAR + MLR · MIMIC-CXR (RADAR) | 14Ma-F1: 0.472 -> 0.475 | Improved |
| add latent-risk mitigation with verification (ablation) | RADAR · MIMIC-CXR (RADAR) | 14Ma-F1: 0.460 -> 0.475 | Improved |
| add dynamic masking (ablation) | RADAR + MEE&MLR · MIMIC-CXR (RADAR) | 14Ma-F1: 0.466 -> 0.490 | Improved |
| use linguistically-aligned report instead of raw GT report (ablation) | RADAR + MEE&MLR · MIMIC-CXR (RADAR) | 14Ma-F1: 0.482 -> 0.490; BLEU-1 0.526 -> 0.522 | Mixed |
| combine explicit-error and latent-risk mitigation (ablation) | RADAR · MIMIC-CXR (RADAR) | 14Ma-F1: 0.482 -> 0.490 | Improved |
| use M2DPO instead of standard DPO (ablation) | R2GenGPT · MIMIC-CXR | 14Ma-F1: 0.420 -> 0.437 | Improved |
| use M2DPO instead of SimPO (ablation) | R2GenGPT · MIMIC-CXR | 14Ma-F1: 0.425 -> 0.437 | Improved |
| use M2DPO instead of standard DPO (ablation) | R2GenGPT · In-house endoscopy | 2F1: 0.832 -> 0.852 | Improved |
| mitigate latent risks (ablation) | RADAR · MIMIC-CXR | mean uncertainty score of correct entities: 0.514 -> 0.282 | Improved |
| apply DPO-Clin instead of MMedPO | R2GenGPT · IU X-Ray | 14Ma-F1: 0.332 -> 0.350 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- + MMedPO — MMedPO: Aligning Medical Vision-Language Models with Clinical-Aware Multimodal Preference Optimization (ICML 2025) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MMedPO: Aligning Medical Vision-Language Models with Clinical-Aware Multimodal Preference Optimization (ICML 2025 · similarity 0.88)
- QoQ-Med: Building Multimodal Clinical Foundation Models with Domain-Aware GRPO Training (NeurIPS 2025 · similarity 0.86)
- PromptMRG: Diagnosis-Driven Prompts for Medical Report Generation (AAAI 2024 · similarity 0.86)
- Radiology Report Generation via Multi-objective Preference Optimization (AAAI 2025 · similarity 0.86)
- DAMPER: A Dual-Stage Medical Report Generation Framework with Coarse-Grained MeSH Alignment and Fine-Grained Hypergraph Matching (AAAI 2025 · similarity 0.86)
- Hierarchical Divide-and-Conquer for Fine-Grained Alignment in LLM-Based Medical Evaluation (AAAI 2025 · 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.
