# Recursive Self-Improvement in Unified Multimodal Models

arXiv: https://arxiv.org/abs/2610.03002 · HTML: https://mlsanity.com/paper/2610.03002/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 25 references that are in the map.
- First posted on arXiv 2 Oct 2026.

## Builds on (most cited first, 23 of 23)

- [STaR: Bootstrapping Reasoning With Reasoning](https://mlsanity.com/paper/2203.14465/) (2022)
- [Chameleon: Mixed-Modal Early-Fusion Foundation Models](https://mlsanity.com/paper/2405.09818/) (2024)
- [Emerging Properties in Unified Multimodal Pretraining](https://mlsanity.com/paper/2505.14683/) (2025)
- [Show-o: One Single Transformer to Unify Multimodal Understanding and Generation](https://mlsanity.com/paper/2408.12528/) (2024)
- [Self-Rewarding Language Models](https://mlsanity.com/paper/2401.10020/) (2024)
- [Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model](https://mlsanity.com/paper/2408.11039/) (2024)
- [Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation](https://mlsanity.com/paper/2410.13848/) (2024)
- [Absolute Zero: Reinforced Self-play Reasoning with Zero Data](https://mlsanity.com/paper/2505.03335/) (2025)
- [Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning](https://mlsanity.com/paper/2507.16746/) (2025)
- [Scaling Text-Rich Image Understanding via Code-Guided Synthetic Multimodal Data Generation](https://mlsanity.com/paper/2502.14846/) (2025)
- [Uni-cot: Towards Unified Chain-of-Thought Reasoning Across Text and Vision](https://mlsanity.com/paper/2508.05606/) (2025)
- [UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision](https://mlsanity.com/paper/2601.03193/) (2026)
- [DataEnvGym: Data Generation Agents in Teacher Environments with Student Feedback](https://mlsanity.com/paper/2410.06215/) (2024)
- [UniRL: Self-Improving Unified Multimodal Models via Supervised and Reinforcement Learning](https://mlsanity.com/paper/2505.23380/) (2025)
- [SRUM: Fine-Grained Self-Rewarding for Unified Multimodal Models](https://mlsanity.com/paper/2510.12784/) (2025)
- [Factuality Matters: When Image Generation and Editing Meet Structured Visuals](https://mlsanity.com/paper/2510.05091/) (2025)
- [BizGenEval: A Systematic Benchmark for Commercial Visual Content Generation](https://mlsanity.com/paper/2603.25732/) (2026)
- [EvolveDirector: Approaching Advanced Text-to-Image Generation with Large Vision-Language Models](https://mlsanity.com/paper/2410.07133/) (2024)
- [SUDER: Self-Improving Unified Large Multimodal Models for Understanding and Generation with Dual Self-Rewards](https://mlsanity.com/paper/2506.07963/) (2025)
- [Ask, Solve, Generate: Self-Evolving Unified Multimodal Understanding and Generation via Self-Consistency Rewards](https://mlsanity.com/paper/2606.27376/) (2026)
- [EVE: Verifiable Self-Evolution of MLLMs via Executable Visual Transformations](https://mlsanity.com/paper/2604.18320/) (2026)
- [DataEvolver: Self-Evolving Multi-Agent Data Construction for Text-Rich Image Generation](https://mlsanity.com/paper/2606.31537/) (2026)
- [Dreaming in Flow: Generative Grounding Feedback for Self-Evolving Unified Multimodal Models](https://mlsanity.com/paper/2609.08282/) (2026)

## Abstract (arXiv)

Unified multimodal models (UMMs) understand and generate both text and images, which lets a model produce its own training data. Existing self-improvement in UMMs keeps supervision on the visual side, where image understanding judges image generation. We propose recursive cross-capability self-improvement (RSI), a training loop in which the text and visual abilities of a UMM supply training data for one another. In each round, the model generates images and reads them to find where it falls short. It then writes programs aimed at these shortcomings, and execution verifies every result against its specification. Verified renders train image generation, while labeled renders and the model's own correct programs train visual understanding and program writing. Program execution thus acts as a source of truth outside the model, so errors do not accumulate across rounds. We study RSI on charts and build BasicChartBench to evaluate open models early in training. On requests worded differently from training, four rounds of RSI raise the score from 45.7% to 60.2%, while continued training stays at 46.3%. Verified construction carries most of the gain, and targeting the model's failures adds 3.5%. Along the way, the share of verified programs rises from 48.9% to 95.2%, and the reader's accuracy on edited renders rises from 55.6% to 87.4%.