UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision

TL;DR

UniCorn introduces a self-improving framework for unified multimodal models that enhances their generative capabilities through self-play and internal supervision, achieving state-of-the-art results without external data.

Contribution

The paper presents UniCorn, a novel self-supervised method that improves multimodal models' generation by partitioning the model into collaborative roles and using self-generated signals, eliminating external supervision.

Findings

Achieves SOTA on multiple image generation benchmarks.

Significantly improves text-to-image generation quality.

Maintains strong multimodal comprehension while enhancing generation.

Abstract

While Unified Multimodal Models (UMMs) have achieved remarkable success in cross-modal comprehension, a significant gap persists in their ability to leverage such internal knowledge for high-quality generation. We formalize this discrepancy as Conduction Aphasia, a phenomenon where models accurately interpret multimodal inputs but struggle to translate that understanding into faithful and controllable synthesis. To address this, we propose UniCorn, a simple yet elegant self-improvement framework that eliminates the need for external data or teacher supervision. By partitioning a single UMM into three collaborative roles: Proposer, Solver, and Judge, UniCorn generates high-quality interactions via self-play and employs cognitive pattern reconstruction to distill latent understanding into explicit generative signals. To validate the restoration of multimodal coherence, we introduce UniCycle, a cycle-consistency benchmark based on a Text to Image to Text reconstruction loop. Extensive experiments demonstrate that UniCorn achieves comprehensive and substantial improvements over the base model across six general image generation benchmarks. Notably, it achieves SOTA performance on TIIF(73.8), DPG(86.8), CompBench(88.5), and UniCycle while further delivering substantial gains of +5.0 on WISE and +6.5 on OneIG. These results highlight that our method significantly enhances T2I generation while maintaining robust comprehension, demonstrating the scalability of fully self-supervised refinement for unified multimodal intelligence.