Research map / Vision-language and multimodal / Cross-modal learning
FlowBind: Efficient Any-to-Any Generation with Bidirectional Flows
Yeonwoo Cha, Semin Kim, Jinhyeon Kwon, Seunghoon Hong
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 |
|---|---|---|---|
| use learnable shared latent anchor instead of fixed text anchor (ablation) | FlowBind w/o I-A · I->T alignment | CLIP: 27.94 -> 30.04 | Improved |
| use learnable shared latent anchor instead of fixed text anchor (ablation) | FlowBind w/o I-A · A->T alignment | CLAP: 36.72 -> 37.04 | Improved |
| use learnable shared latent anchor instead of fixed text anchor (ablation) | FlowBind w/o I-A · I->A alignment | AIS: 55.48 -> 61.88 | Improved |
| compare shared latent against per-modality latent (ablation) | FlowBind · text-audio CKNNA | CKNNA: 0.1965 -> 0.2872 | Improved |
| compare shared latent against per-modality latent (ablation) | FlowBind · audio-image CKNNA | CKNNA: 0.1343 -> 0.3026 | Improved |
| FlowBind versus OmniFlow on image-audio generation quality | FlowBind · I->A | FAD: 5.67 -> 2.50 | Improved |
| FlowBind versus OmniFlow on text-to-image alignment | FlowBind · T->I (MS-COCO) | CLIP: 31.52 -> 28.35 | Hurt |
| FlowBind versus OmniFlow on audio-to-text alignment | FlowBind · A->T | CLAP: 45.08 -> 36.70 | Hurt |
| FlowBind versus OmniFlow on (T+I)->A alignment | FlowBind · Many-to-one | CLAP: 7.68 -> 28.13 | Improved |
| FlowBind versus OmniFlow on T->(I+A) generation | FlowBind · One-to-many | CLAP (T->A): 12.92 -> 29.12 | Improved |
| FlowBind versus specialist TangoFlux on text-to-audio | FlowBind · T->A | FAD: 1.41 -> 4.19 | Hurt |
Closest papers by meaning
- OmniFlow: Any-to-Any Generation with Multi-Modal Rectified Flows (CVPR 2025 · similarity 0.88)
- Flowing from Words to Pixels: A Noise-Free Framework for Cross-Modality Evolution (CVPR 2025 · similarity 0.88)
- Towards General Modality Translation with Contrastive and Predictive Latent Diffusion Bridge (NeurIPS 2025 · similarity 0.86)
- UniFlow-Audio: Unified Flow Matching for Audio Generation from Omni-Modalities (NeurIPS 2026 · similarity 0.86)
- Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data (ICLR 2024 · similarity 0.85)
- X-AVDD: Cross-Attentive Audio-Visual Dataset Distillation (NeurIPS 2026 · similarity 0.85)
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.
