AI Paper Atlas

Research map / Benchmarks / MIMIC-CXR

MIMIC-CXR leaderboard

30 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MIMIC-CXR. Best BLEU-4: 0.193 (DAMPER, AAAI 2025).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

MIMIC-CXR · BLEU-4

25 papers, each with its best reported number for its own method; higher is better.

#BLEU-4 ↑MethodPaperTopic
10.193DAMPERDAMPER: A Dual-Stage Medical Report Generation Framework with Coarse-Grained MeSH Alignment and Fine-Grained Hypergraph Matching
AAAI 2025 · Table 1
Human and embodied modeling
20.182REVTAF (Ours)Learnable Retrieval Enhanced Visual-Text Alignment and Fusion for Radiology Report Generation
ICCV 2025 · Table 1
Cross-modal learning
30.175PriorRGPriorRG: Prior-Guided Contrastive Pre-training and Coarse-to-Fine Decoding for Chest X-ray Report Generation
AAAI 2026 · Table 1
Human and embodied modeling
40.164RefleXNet
Qwen2.5-VL 3B
RefleXNet: Targeted Self-Reflection for Accurate Chest X-ray Reporting
AAAI 2026 · Table 1
Multimodal LLMs
50.158MLRGEnhanced Contrastive Learning with Multi-view Longitudinal Data for Chest X-ray Report Generation
CVPR 2025 · Table 2
3D generation
60.155CPO
Qwen2.5-VL (7B)
Walking the Tightrope: Autonomous Disentangling Beneficial and Detrimental Drifts in Non-Stationary Custom-Tuning
NeurIPS 2025 · Table 2
Instruction and fine-tuning
70.143Ours
LLaMA3-3B + Swin
SAT-RRG: LLM-Guided Self-Adaptive Training for Radiology Report Generation with Token-Level Push-Pull Optimization
CVPR 2026 · Table 2
Language, safety and interpretability
80.143REVA-PO
Qwen2.5-VL-7B-Instruct
REVA-PO: Stabilizing Reinforcement Learning for Chest X-ray Report Generation
ECCV 2026 · Table 1
RL from feedback
90.137DARTDART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation
CVPR 2025 · Table 1
Text-to-image diffusion
100.134DiADiA-gnostic VLVAE: Disentangled Alignment-Constrained Vision Language Variational AutoEncoder for Robust Radiology Reporting with Missing Modalities
AAAI 2026 · Table 1
Cross-modal learning
15 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

MIMIC-CXR · ROUGE-L

18 papers, each with its best reported number for its own method; higher is better.

#ROUGE-L ↑MethodPaperTopic
10.369DiADiA-gnostic VLVAE: Disentangled Alignment-Constrained Vision Language Variational AutoEncoder for Robust Radiology Reporting with Missing Modalities
AAAI 2026 · Table 1
Cross-modal learning
20.324PriorRGPriorRG: Prior-Guided Contrastive Pre-training and Coarse-to-Fine Decoding for Chest X-ray Report Generation
AAAI 2026 · Table 1
Human and embodied modeling
30.321CPO
Qwen2.5-VL (7B)
Walking the Tightrope: Autonomous Disentangling Beneficial and Detrimental Drifts in Non-Stationary Custom-Tuning
NeurIPS 2025 · Table 2
Instruction and fine-tuning
40.31DARTDART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation
CVPR 2025 · Table 1
Text-to-image diffusion
50.309RefleXNet
Qwen2.5-VL 3B
RefleXNet: Targeted Self-Reflection for Accurate Chest X-ray Reporting
AAAI 2026 · Table 1
Multimodal LLMs
60.309MPO
R2Gen
Radiology Report Generation via Multi-objective Preference Optimization
AAAI 2025 · NLG comparison table
RL from feedback
70.309MedM2GMedM2G: Unifying Medical Multi-Modal Generation via Cross-Guided Diffusion with Visual Invariant
CVPR 2024 · Table 1
Cross-modal learning
80.304CoFE
ViT-S + GPT-2 Medium
Learning with Counterfactual Explanations for Radiology Report Generation
ECCV 2024 · Table 1
Human and embodied modeling
90.303Ours
LLaMA3-3B + Swin
SAT-RRG: LLM-Guided Self-Adaptive Training for Radiology Report Generation with Token-Level Push-Pull Optimization
CVPR 2026 · Table 2
Language, safety and interpretability
100.301DAMPERDAMPER: A Dual-Stage Medical Report Generation Framework with Coarse-Grained MeSH Alignment and Fine-Grained Hypergraph Matching
AAAI 2025 · Table 1
Human and embodied modeling
8 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

MIMIC-CXR · BLEU-1

14 papers, each with its best reported number for its own method; higher is better.

#BLEU-1 ↑MethodPaperTopic
10.465REVTAF (Ours)Learnable Retrieval Enhanced Visual-Text Alignment and Fusion for Radiology Report Generation
ICCV 2025 · Table 1
Cross-modal learning
20.426CPO
Qwen2.5-VL (7B)
Walking the Tightrope: Autonomous Disentangling Beneficial and Detrimental Drifts in Non-Stationary Custom-Tuning
NeurIPS 2025 · Table 2
Instruction and fine-tuning
30.419EKAGen (RN-101)
ResNet-101
Instance-level Expert Knowledge and Aggregate Discriminative Attention for Radiology Report Generation
CVPR 2024 · Table 1
Recognition and long-tail
40.416MPO
R2Gen
Radiology Report Generation via Multi-objective Preference Optimization
AAAI 2025 · NLG comparison table
RL from feedback
50.415DiADiA-gnostic VLVAE: Disentangled Alignment-Constrained Vision Language Variational AutoEncoder for Robust Radiology Reporting with Missing Modalities
AAAI 2026 · Table 1
Cross-modal learning
60.412PriorRGPriorRG: Prior-Guided Contrastive Pre-training and Coarse-to-Fine Decoding for Chest X-ray Report Generation
AAAI 2026 · Table 1
Human and embodied modeling
70.402DAMPERDAMPER: A Dual-Stage Medical Report Generation Framework with Coarse-Grained MeSH Alignment and Fine-Grained Hypergraph Matching
AAAI 2025 · Table 1
Human and embodied modeling
80.402MiniGPT-4 (FT)+I3+C2FD (THIS WORK)
MiniGPT-4 (FT)
Bootstrapping Large Language Models for Radiology Report Generation
AAAI 2024 · Table 2
Language, safety and interpretability
90.398PromptMRG
ResNet-101 + BERT-base
PromptMRG: Diagnosis-Driven Prompts for Medical Report Generation
AAAI 2024 · Table 1
Human and embodied modeling
100.396Ours
DenseNet121; BERT-base
Automatic Radiology Reports Generation via Memory Alignment Network
AAAI 2024 · Table 1
Human and embodied modeling
4 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

MIMIC-CXR · METEOR

14 papers, each with its best reported number for its own method; higher is better.

#METEOR ↑MethodPaperTopic
10.289DAMPERDAMPER: A Dual-Stage Medical Report Generation Framework with Coarse-Grained MeSH Alignment and Fine-Grained Hypergraph Matching
AAAI 2025 · Table 1
Human and embodied modeling
20.236CPO
Qwen2.5-VL (7B)
Walking the Tightrope: Autonomous Disentangling Beneficial and Detrimental Drifts in Non-Stationary Custom-Tuning
NeurIPS 2025 · Table 2
Instruction and fine-tuning
30.199REVTAF (Ours)Learnable Retrieval Enhanced Visual-Text Alignment and Fusion for Radiology Report Generation
ICCV 2025 · Table 1
Cross-modal learning
40.189PriorRGPriorRG: Prior-Guided Contrastive Pre-training and Coarse-to-Fine Decoding for Chest X-ray Report Generation
AAAI 2026 · Table 1
Human and embodied modeling
50.176CoFE
ViT-S + GPT-2 Medium
Learning with Counterfactual Explanations for Radiology Report Generation
ECCV 2024 · Table 1
Human and embodied modeling
60.168RGRG
GPT-2 Medium
Interactive and Explainable Region-Guided Radiology Report Generation
CVPR 2023 · Table 1
Human and embodied modeling
70.167Ours
LLaMA3-3B + Swin
SAT-RRG: LLM-Guided Self-Adaptive Training for Radiology Report Generation with Token-Level Push-Pull Optimization
CVPR 2026 · Table 2
Language, safety and interpretability
80.161MMTNMMTN: Multi-Modal Memory Transformer Network for Image-Report Consistent Medical Report Generation
AAAI 2023 · Table 1
Human and embodied modeling
90.16KiUTKiUT: Knowledge-Injected U-Transformer for Radiology Report Generation
CVPR 2023 · Table 1
Human and embodied modeling
100.157UAR
Transformer-Base (dVAE)
Unify, Align and Refine: Multi-Level Semantic Alignment for Radiology Report Generation
ICCV 2023 · Table 1
Cross-modal learning
4 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

MIMIC-CXR · F1

11 papers, each with its best reported number for its own method; higher is better.

#F1 ↑MethodPaperTopic
10.533DARTDART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation
CVPR 2025 · Table 2
Text-to-image diffusion
20.507DAMPERDAMPER: A Dual-Stage Medical Report Generation Framework with Coarse-Grained MeSH Alignment and Fine-Grained Hypergraph Matching
AAAI 2025 · Table 2
Human and embodied modeling
30.505MLRGEnhanced Contrastive Learning with Multi-view Longitudinal Data for Chest X-ray Report Generation
CVPR 2025 · Table 2
3D generation
40.499EKAGen (RN-101)
ResNet-101
Instance-level Expert Knowledge and Aggregate Discriminative Attention for Radiology Report Generation
CVPR 2024 · Table 2
Recognition and long-tail
50.476PromptMRG
ResNet-101 + BERT-base
PromptMRG: Diagnosis-Driven Prompts for Medical Report Generation
AAAI 2024 · Table 1
Human and embodied modeling
60.473OURS (14.2B)
MiniGPT-4 (FT)
Bootstrapping Large Language Models for Radiology Report Generation
AAAI 2024 · Table 3
Language, safety and interpretability
70.389Ours
DenseNet121; BERT-base
Automatic Radiology Reports Generation via Memory Alignment Network
AAAI 2024 · Table 2
Human and embodied modeling
80.373DCLDynamic Graph Enhanced Contrastive Learning for Chest X-Ray Report Generation
CVPR 2023 · Table 2
Knowledge and temporal graphs
90.353MPO
R2Gen
Radiology Report Generation via Multi-objective Preference Optimization
AAAI 2025 · Clinical metrics table
RL from feedback
100.321KiUTKiUT: Knowledge-Injected U-Transformer for Radiology Report Generation
CVPR 2023 · Table 1
Human and embodied modeling
1 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.