[Submitted on 8 Sep 2026 (v1), last revised 25 Oct 2026 (this version, v2)]
Abstract:Unified multimodal models can generate images and interpret them, but their two capabilities are usually optimized with separate supervision. We introduce Bidirectional Self-Refinement via Flow-Direction Feedback and Semantic Replay (BiSe), a post-training method that uses self-generated visual episodes to construct feedback for both pathways. Understanding-guided generation aligns the prompt-conditioned flow direction with an image-conditioned prediction and with the model’s response to a repair instruction. Generation-guided understanding replays the source intent through captioning and image generation, and learns a stable semantic evidence profile across the replayed images while separating an independent episode. Flow-matching and source-prompt objectives remain active, and frozen initial-model predictions limit drift. BiSe requires no paired image–text data during post-training. It improves text-to-image compositionality across three unified backbones and visual understanding on the two backbones evaluated for that task.
Submission history
From: Ke Hao [view email]
[v1] Tue, 8 Sep 2026 05:52:53 UTC (1,626 KB)
[v2] Sun, 25 Oct 2026 15:46:22 UTC (18,869 KB)
