Research map / Vision-language and multimodal / Human and embodied modeling
Learning with Counterfactual Explanations for Radiology Report Generation
Mingjie Li, Haokun Lin, Liang Qiu, Xiaodan Liang, Ling Chen, Abdulmotaleb Elsaddik, Xiaojun Chang
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 pre-trained PubMedBERT and GPT-2 Medium as text encoder and decoder (ablation) | BLIP base · IU-Xray | CIDEr: 0.363 -> 0.510 | Improved |
| add counterfactual contrastive loss L_CF (ablation) | LLM-based model · IU-Xray | CIDEr: 0.510 -> 0.678 | Improved |
| add counterfactual loss on top of LLM and prompt (ablation) | LLM + prompt · IU-Xray | CIDEr: 0.698 -> 0.731 | Improved |
| add learnable factual and counterfactual prompt (ablation) | LLM-based model · IU-Xray | CIDEr: 0.510 -> 0.698 | Improved |
| add learnable prompt (ablation) | LLM-based model · MIMIC-CXR clinical efficacy | F1-score: 0.317 -> 0.366 | Improved |
| add L_CF to full model for clinical metrics (ablation) | LLM + prompt · MIMIC-CXR clinical efficacy | F1-score: 0.366 -> 0.405 | Improved |
| use BLEU-based hard negative sampling instead of random sampling (ablation) | CoFE · IU-Xray | CIDEr: 0.653 -> 0.678 | Improved |
| use BLEU-based negative sampling in the full model (ablation) | CoFE · IU-Xray | CIDEr: 0.706 -> 0.731 | Improved |
| CoFE compared with prior methods on MIMIC-CXR NLG | CoFE · MIMIC-CXR | CIDEr: CIDEr 0.453 vs RGRG 0.495; ROUGE-L 0.304 | Mixed |
Closest papers by meaning
- Learnable Retrieval Enhanced Visual-Text Alignment and Fusion for Radiology Report Generation (ICCV 2025 · similarity 0.88)
- AHIVE: Anatomy-aware Hierarchical Vision Encoding for Interactive Radiology Report Retrieval (CVPR 2024 · similarity 0.86)
- S2D-Align: Shallow-to-Deep Auxiliary Learning for Anatomically-Grounded Radiology Report Generation (AAAI 2026 · similarity 0.86)
- Bootstrapping Large Language Models for Radiology Report Generation (AAAI 2024 · similarity 0.86)
- DART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation (CVPR 2025 · similarity 0.85)
- Automatic Radiology Reports Generation via Memory Alignment Network (AAAI 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.
