AI Paper Atlas

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

Scientific Image Synthesis: Benchmarking, Methodologies, and Downstream Utility

honglin lin, Chonghan Qin, Zheng Liu, Qizhi Pei, Yu Li, Zhanping Zhong, Xin Gao, Wei Li, Wentao Zhang, Yanfeng Wang, Conghui He, Lijun Wu

ECCV 2026 · Image generation › Text-to-image diffusion · approach: T2I evaluation and controllable adapters · 15 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
generate science images via code (ImgCoder) instead of pixel T2I with the same Qwen familyQwen3-ImgCoder vs Qwen-Image · SciGenBenchRinv (%): 38.86 -> 56.38Improved
use Gemini-3-Pro-ImgCoder instead of best pixel model Nanobanana-Proscientific image generation · SciGenBenchRinv (%): 73.41 -> 77.87Improved
code-driven vs pixel-based generation on biologyQwen3-ImgCoder · SciGenBench BiologyRinv (%): Qwen-Image 43.14 vs Qwen3-ImgCoder 41.18Hurt
code-driven vs pixel-based generation on image fidelity metricsImgCoder · SciGenBench-SeePhysFID: GPT-Image-1 77.31 vs Gemini-3-Pro-ImgCoder 107.67Hurt
RL fine-tune LMM on verified synthetic scientific imagesQwen3-VL-8B-Instruct · GEO3K and MathVision averageaverage accuracy: 54.5 -> 58.2Improved
filter incorrect synthetic images from training data (ablation)Qwen-Image data · GEO3K+MathVision averageaverage accuracy: 57.1 -> 57.8Improved
filter incorrect synthetic images from training data (ablation)Nanobanana-Pro data · GEO3K / MathVisionaccuracy: GEO3K 70.1 -> 68.7; MV 46.1 -> 47.7; AVG 58.1 -> 58.2Mixed
use higher-quality generator (Nanobanana-Pro instead of Qwen-Image) for training images (ablation)RL data · GEO3K+MathVision averageaverage accuracy: 57.1 -> 58.1Improved
scale synthetic training set from 50 to 1.4K samples (ablation)Nanobanana-Pro images · downstream accuracyaccuracy: 43.9 -> 46.1Improved
use code-generated (ImgCoder) vs pixel images for training (ablation)Qwen3-VL-8B RL · MathVisionaccuracy: Qwen-Image 45.9 -> Qwen-Imgcoder 46.9Improved
use ImgCoder vs pixel on GEO3K training data (ablation)Qwen3-VL-8B RL · GEO3Kaccuracy: Qwen-Image 68.2 vs Qwen-Imgcoder 67.9No clear effect

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