arXiv cs.AI· Shuo Han, Zhilin Wang, Peng Ling, Wenjia Gong, Yang Liu, Guanjun Jiang, Yukun Cao, Xike Xie·· 12 小时前AI 评分34
ImageRSI：面向智能体图像生成的验证驱动递归自我改进框架
ImageRSI: Verification-Driven Recursive Self-Improvement for Agentic Image Generation
AI 导读
ImageRSI 提出一种保持图像生成器冻结的框架，通过任务内五维验证与修复、跨任务经验库和离线更新评判器实现递归自我改进，经验需经配对 A/B 验证才能入库。
正文 · 原文
Abstract:Complex text-to-image requests remain difficult to satisfy reliably. Verifying and repairing an image can fix an individual output, but a successful repair does not necessarily carry over to other requests. We present ImageRSI, a framework that keeps the image generator frozen and combines three components: five-dimensional verification and repair within each task, an experience library shared across tasks, and offline updating of the judge. Candidate experiences enter the library only after paired A/B validation, and the judge is updated with supervised fine-tuning and GRPO on accumulated image-judgment records and evaluated before reuse. We also introduce ImageReq-Bench, a benchmark of 961 requests built around hard cases in real-world image-generation requests, each paired with an evaluation checklist. ImageRSI achieves a Checklist Strict Accuracy (CSA) of 52.6% on MindBench (with Nano Banana Pro as the generator) and 65.66% on ImageReq-Bench, and we compare system outcomes across judge-training configurations.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.32120 [cs.CV] |
| (or arXiv:2610.32120v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.32120 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Shuo Han [view email]
[v1]
Fri, 23 Oct 2026 03:42:21 UTC (2,414 KB)
来源：arXiv cs.AI · arxiv.org