Research map / Image generation / Text-to-image diffusion
SciIR: A Large-scale Training Dataset and Benchmark for Scientific Image Reasoning Generation
Zhiyuan Ma, Zhengfeng Shi, Yuning An, Peize Li, Jiabao Wei, Ruijie Li, Junhao Xiao, Jianjun Li, Bowen Zhou
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 |
|---|---|---|---|
| fine-tune Qwen-Image-2512 on SciIR-82k with reasoning planner | Qwen-Image-2512 · SciIR-Bench | Final score: 35 -> 43 | Improved |
| add Sci-RCoT training data (ablation) | Qwen-Image-SciIR · SciIR-Bench | Final score: 38 -> 43 | Improved |
| add reasoning planner (ablation) | Qwen-Image-SciIR · SciIR-Bench | Final score: 41 -> 43 | Improved |
| add taxonomy (ablation) | Qwen-Image-SciIR · SciIR-Bench | Final score: 39 -> 43 | Improved |
| add reasoning planner (ablation) | Qwen-Image-SciIR · SciIR-Bench | Scientific Process: 49 -> 53 | Improved |
| add Sci-RCoT (ablation) | Qwen-Image-SciIR · SciIR-Bench | Scientific Process: 39 -> 53 | Improved |
| fine-tune on SciIR-82k | Qwen-Image-2512 · SciIR-Bench | Scientific Process: 37 -> 53 | Improved |
| fine-tune on SciIR-82k | Qwen-Image-2512 · SciIR-Bench | Entity Structure: 50 -> 59 | Improved |
| fine-tune on SciIR-82k | Qwen-Image-2512 · SciIR-Bench | Text: 15 -> 15 | No clear effect |
| fine-tune on SciIR-82k | Qwen-Image-2512 · SciIR-Bench | Scientific Law avg: 40 -> 43 | Improved |
| use dense Sci-RCoT prompt instead of abstract prompt (IF vs IR) | FLUX.1-Kontext-Max · SciIR-Bench | Scientific Process: 13 -> 36 | Improved |
| use dense Sci-RCoT prompt instead of abstract prompt (IF vs IR) | Flux-Dev · SciIR-Bench | Scientific Law: 25 -> 6 | Hurt |
Closest papers by meaning
- MMMG: A Massive, Multidisciplinary, Multi-Tier Generation Benchmark for Text-to-Image Reasoning (NeurIPS 2025 · similarity 0.90)
- ScImage: How good are multimodal large language models at scientific text-to-image generation? (ICLR 2025 · similarity 0.90)
- GIR-Bench: Versatile Benchmark for Generating Images with Reasoning (ICLR 2026 · similarity 0.88)
- MILR: Improving Multimodal Image Generation via Test-Time Latent Reasoning (ICLR 2026 · similarity 0.88)
- Factuality Matters: When Image Generation and Editing Meet Structured Visuals (ICLR 2026 · similarity 0.87)
- SciVideoBench: Benchmarking Scientific Video Reasoning in Large Multimodal Models (ICML 2026 · similarity 0.87)
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.
