AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

SC-Captioner: Improving Image Captioning with Self-Correction by Reinforcement Learning

Lin Zhang, Xianfang Zeng, Kangcong Li, Gang Yu, Tao Chen

ICCV 2025 · Image generation › Text-to-image diffusion · approach: T2I evaluation and controllable adapters · 11 citations (OpenAlex)

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 SFT on RefinedCapsQwen2-VL-7B · DOCCI500CAPTURE: 57.96 -> 62.05Improved
add self-correction RL (ours) after SFT instead of SFT aloneQwen2-VL-7B · DOCCI500CAPTURE: 62.05 -> 63.34Improved
add self-correction RL (ours) instead of DPO after SFTQwen2-VL-7B · DOCCI500Relations QA: 27.33 -> 30.51Improved
add self-correction RL (ours) after SFTLLaVA-1.5-7B · DOCCI500CAPTURE: 60.92 -> 62.29Improved
add DPO self-correction after SFTQwen2-VL-7B · DOCCI500Objects precision: 79.99 -> 78.78Hurt
add prompting-only self-correction to SFT modelQwen2-VL-7B · DOCCI500CAPTURE: 62.05 -> 62.02No clear effect
use mistake punishment in the reward (ablation)SC-Captioner reward · DOCCI500Objects precision: 77.52 -> 80.20Improved
use correctness bonus and mistake punishment instead of CAPTURE score as reward (ablation)SC-Captioner reward · DOCCI500CAPTURE: 62.34 -> 63.34Improved
add object reward component only (ablation)SC-Captioner reward · DOCCI500Objects F1: 69.32 -> 71.44Improved
reward from second-turn correction instead of first turn only (ablation)SC-Captioner · DOCCI500CAPTURE: 63.12 -> 63.34Improved
use CLIP-based DiscriTune instead of correction reward (ablation)SFT Qwen2-VL-7B · DOCCI500CAPTURE: 62.05 -> 60.37Hurt
train on RefinedCaps instead of DOCCI train set (ablation)Qwen2-VL-7B · COCO-LN500 (cross-domain)Relations QA: 32.01 -> 38.51Improved

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