Research map / Vision-language and multimodal / Human and embodied modeling
FlowInOne: Unifying Multimodal Generation as Image-in, Image-out Flow Matching
Junchao Yi, Rui Zhao, Jiahao Tang, Weixian Lei, Linjie Li, Qisheng Su, Zhengyuan Yang, Lijuan Wang, Xiaofeng Zhu, Alex Jinpeng Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train jointly on all 5M samples instead of two-stage (ablation) | FlowInOne · VP-Bench | avg pass rate: 29.1% -> 47.8% | Improved |
| add dual-path cross-attention to source latent (ablation) | FlowInOne · VP-Bench | avg pass rate: 0.182 -> 0.214 | Improved |
| add spatially-adaptive gating to dual-path cross-attention (ablation) | FlowInOne · VP-Bench | avg pass rate: 0.214 -> 0.231 | Improved |
| use MLP+MLP compression instead of truncation (ablation) | FlowInOne · VP-Bench | avg pass rate: 0.169 -> 0.182 | Improved |
| use MLP+MLP compression instead of VAE expansion (ablation) | FlowInOne · VP-Bench | avg pass rate: 0.157 -> 0.182 | Improved |
| render instructions into image canvas (image-in image-out) vs text-conditioned open-source editors | FlowInOne · VP-Bench force understanding (Gemini3) | success rate: .727 vs .047 (Qwen-IE-2509) | Improved |
| FlowInOne vs commercial Nano Banana on total success | FlowInOne · VP-Bench | total success rate: .540 vs .589 (Gemini3); .392 vs .391 (GPT5.2) | Mixed |
| FlowInOne vs Nano Banana on visual realism | FlowInOne · VP-Bench (Gemini3) | visual realism: 4.43 -> 3.12 | Hurt |
| remove in-image text, keep visual prompts only (ablation) | FlowInOne · visual instruction ablation (4 samples) | qualitative | Hurt |
| replace valid instruction text with random corrupted text (ablation) | FlowInOne · robustness test | qualitative | Hurt |
| lower input resolution to 128x128 (ablation) | FlowInOne · robustness test | qualitative | Hurt |
| change text style/length/blur of rendered instruction (ablation) | FlowInOne · robustness test | qualitative | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Omnigen2 — OmniGen2: Towards Instruction-Aligned Multimodal Generation (CVPR 2026) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation (NeurIPS 2026 · similarity 0.89)
- Best of Both Worlds: Multimodal Reasoning and Generation via Unified Discrete Flow Matching (ICML 2026 · similarity 0.89)
- DuoGen: Towards Autonomous Interleaved Multimodal Generation (CVPR 2026 · similarity 0.88)
- STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation (NeurIPS 2026 · similarity 0.88)
- Vision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations (NeurIPS 2025 · similarity 0.88)
- FUDOKI: Discrete Flow-based Unified Understanding and Generation via Kinetic-Optimal Velocities (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
