Research map / Optimization / Optimizers
The Double Dilemma in Multi-Task Radiology Report Generation: A Gradient Dynamics Analysis and Solution
Erjian Zhang, Yatong Hao, Liejun Wang, Zhiqing Guo
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 |
|---|---|---|---|
| replace linear scalarization with CAME-Grad | REVTAF · MIMIC-CXR test | CE average: 0.607 -> 0.626 | Improved |
| replace linear scalarization with CAME-Grad | DCL · MIMIC-CXR test | CE average: 0.249 -> 0.301 | Improved |
| replace linear scalarization with CAME-Grad | PromptMRG · MIMIC-CXR test | B-1: 0.396 -> 0.396 | No clear effect |
| replace linear scalarization with CAME-Grad | DDaTR · IU X-Ray test (zero-shot) | CE average: 0.282 -> 0.283 | No clear effect |
| replace linear scalarization with CAME-Grad | WCL · IU X-Ray test (supervised) | CE average: 0.493 -> 0.517 | Improved |
| add conflict-averse direction rectification (S1) before magnitude enhancement (ablation) | CAME-Grad on PromptMRG · MIMIC-CXR test | F1: 0.476 -> 0.491 | Improved |
| add magnitude-enhanced energy injection (S2) (ablation) | CAME-Grad on DDaTR · MIMIC-CXR test | F1: 0.425 -> 0.438 | Improved |
| add adaptive gradient fusion (S3) (ablation) | CAME-Grad on DDaTR · MIMIC-CXR test | precision: 0.425 -> 0.432 | Improved |
| add adaptive gradient fusion (S3) (ablation) | CAME-Grad on PromptMRG · MIMIC-CXR test | recall: 0.531 -> 0.530 | No clear effect |
| use CAME-Grad instead of CAGrad on REVTAF | REVTAF · MIMIC-CXR test | CE average: 0.615 -> 0.626 | Improved |
| use MMPareto instead of linear scalarization | REVTAF · MIMIC-CXR test | CE average: 0.607 -> 0.076 | Hurt |
| use GradNorm instead of linear scalarization | REVTAF · MIMIC-CXR test | CE average: 0.607 -> 0.604 | No clear effect |
Closest papers by meaning
- Radiology Report Generation via Multi-objective Preference Optimization (AAAI 2025 · similarity 0.89)
- ReFINE: A Reward-Based Framework for Interpretable and Nuanced Evaluation of Radiology Report Generation (AAAI 2026 · similarity 0.87)
- LLM-RG4: Flexible and Factual Radiology Report Generation Across Diverse Input Contexts (AAAI 2025 · similarity 0.87)
- Bootstrapping Large Language Models for Radiology Report Generation (AAAI 2024 · similarity 0.86)
- S2D-Align: Shallow-to-Deep Auxiliary Learning for Anatomically-Grounded Radiology Report Generation (AAAI 2026 · similarity 0.86)
- GEMA-Score: Granular Explainable Multi-Agent Scoring Framework for Radiology Report Evaluation (AAAI 2026 · similarity 0.86)
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.
