DataEvolver: Self-Evolving Multi-Agent Data Construction for Text-Rich Image Generation

Hugging Face Daily Papers Papers

Summary

DataEvolver is a self-evolving multi-agent framework that leverages feedback from rejected samples to iteratively enhance data quality for text-rich image generation, achieving 85.3% OCR-F1 improvement on TextScenesHQ and 35.3% on LongTextBench using PixArt-alpha.

Text-rich image generation is one of the most challenging settings in image generation, since models must simultaneously produce visually realistic images and render legible, semantically aligned, and layout-consistent text. Existing data pipelines usually follow a static crawl-filter-freeze paradigm. They collect candidate samples, filter them once, and freeze the accepted data for training. However, rejected samples are usually discarded, although they often contain useful failure signals such as OCR errors and semantic mismatches. As a result, later construction rounds may repeat the same failure modes. To address these limitations, we propose DataEvolver, a self-evolving multi-agent framework for text-rich image data construction. DataEvolver treats data construction as feedback-driven construction policy evolution. A Retriever collects candidate samples, a Verifier assigns quality scores and rejection causes, a Critic summarizes round-level feedback into semantic feedback, and a Generator completes under-covered regions through targeted synthesis. The updated feedback memory then guides the next construction round. Experiments on text-rich image generation benchmarks show that DataEvolver produces more useful training data than fixed-dataset baselines under matched data budgets. At the 0.75M scale on PixArt-alpha, DataEvolver improves OCR-F1 over the strongest baseline by 85.3 percent on TextScenesHQ and 35.3 percent on LongTextBench. The improvements are consistent across both evaluated benchmarks and also transfer to Show-o2, indicating that the benefit of DataEvolver is not tied to a single downstream generator. These results suggest that rejected samples can provide actionable feedback for improving text-rich image data construction.
Original Article
View Cached Full Text

Cached at: 07/01/26, 11:42 AM

Paper page - DataEvolver: Self-Evolving Multi-Agent Data Construction for Text-Rich Image Generation

Source: https://huggingface.co/papers/2606.31537

Abstract

DataEvolver is a self-evolving multi-agent framework that improves text-rich image generation by leveraging feedback from rejected samples to iteratively enhance data quality.

Text-rich image generationis one of the most challenging settings in image generation, since models must simultaneously produce visually realistic images and render legible, semantically aligned, and layout-consistent text. Existing data pipelines usually follow a static crawl-filter-freeze paradigm. They collect candidate samples, filter them once, and freeze the accepted data for training. However, rejected samples are usually discarded, although they often contain useful failure signals such as OCR errors and semantic mismatches. As a result, later construction rounds may repeat the same failure modes. To address these limitations, we propose DataEvolver, a self-evolvingmulti-agent frameworkfor text-rich imagedata construction. DataEvolver treatsdata constructionas feedback-driven construction policy evolution. A Retriever collects candidate samples, a Verifier assigns quality scores and rejection causes, a Critic summarizes round-level feedback intosemantic feedback, and a Generator completes under-covered regions through targeted synthesis. The updated feedback memory then guides the next construction round. Experiments ontext-rich image generationbenchmarks show that DataEvolver produces more useful training data than fixed-dataset baselines under matcheddata budgets. At the 0.75M scale on PixArt-alpha, DataEvolver improvesOCR-F1over the strongest baseline by 85.3 percent on TextScenesHQ and 35.3 percent on LongTextBench. The improvements are consistent across both evaluated benchmarks and also transfer to Show-o2, indicating that the benefit of DataEvolver is not tied to a singledownstream generator. These results suggest that rejected samples can provide actionable feedback for improving text-rich imagedata construction.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2606\.31537

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2606.31537 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.31537 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.31537 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

EvoMap/evolver

GitHub Trending (daily)

Evolver is a GEP-powered self-evolution engine for AI agents that automates prompt optimization and creates auditable, reusable evolution assets. The project is transitioning from fully open source to source-available while maintaining backward compatibility with existing MIT and GPL-3.0 releases.

CoEvolve: Training LLM Agents via Agent-Data Mutual Evolution

arXiv cs.CL

CoEvolve proposes an agent-data mutual evolution framework for training LLM agents through closed-loop, interaction-driven learning that adapts both the agent and its training data distribution. The method extracts feedback signals from rollout trajectories to guide LLM-based task synthesis, demonstrating significant improvements (15-19% absolute gains) across multiple Qwen models on AppWorld and BFCL benchmarks.

DALL·E: Creating images from text

OpenAI Blog

OpenAI introduces DALL·E, a 12-billion parameter transformer model that generates images from text descriptions by treating text and images as a single token stream. The model demonstrates diverse capabilities including creating anthropomorphized objects, combining disparate concepts, rendering text, and performing image inpainting tasks.