AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

SAT-RRG: LLM-Guided Self-Adaptive Training for Radiology Report Generation with Token-Level Push-Pull Optimization

Yunyi Liu, Yingshu Li, Tong Chen, Lingqiao Liu, Lei Wang, Luping Zhou

CVPR 2026 · Large language models › Language, safety and interpretability · approach: Data and evaluation methodology

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 changedWhereMeasuredEffect
add ETAPL error-token penalty loss (ablation)CE baseline · MIMIC-CXRBLEU-4: 0.131 -> 0.136Improved
add CTAL correct-token loss (ablation)CE baseline · MIMIC-CXRBLEU-4: 0.131 -> 0.141Improved
add ETAPL on top of CTAL (ablation)CTAL · MIMIC-CXRBLEU-4: 0.141 -> 0.143Improved
add focal weighting (ablation)entropy-only SAT-RRG · MIMIC-CXRBLEU-4: 0.139 -> 0.143Improved
add entropy weighting (ablation)focal-only SAT-RRG · MIMIC-CXRBLEU-4: 0.141 -> 0.143Improved
increase focal gamma from 1.5 to 2.0 (ablation)SAT-RRG · MIMIC-CXRBLEU-4: 0.143 -> 0.141Hurt
decrease focal gamma from 1.5 to 1.0 (ablation)SAT-RRG · MIMIC-CXRBLEU-4: 0.143 -> 0.139Hurt
train with SAT-RRG token-level push-pull lossesLLaMA3-3B report generator · IU-XrayBLEU-4: 0.190 -> 0.196 vs EKAGenImproved
SAT-RRG vs EKAGen on RadCliQSAT-RRG · MIMIC-CXRRadCliQ: 1.126 -> 1.150Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy