Research map / Vision-language and multimodal / Cross-modal learning
CaReFlow: Cyclic Adaptive Rectified Flow for Multimodal Fusion
Sijie Mai, Shiqin Han
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 |
|---|---|---|---|
| add rectified-flow modality distribution alignment (ablation) | CaReFlow · CMU-MOSI | Acc7: 45.9 -> 50.6 | Improved |
| add rectified-flow modality distribution alignment (ablation) | CaReFlow · CH-SIMS-v2 | Acc2: 78.4 -> 82.9 | Improved |
| add cyclic information flow (backward rectified flow) (ablation) | CaReFlow · CH-SIMS-v2 | Acc2: 81.7 -> 82.9 | Improved |
| add cyclic information flow (ablation) | CaReFlow · CMU-MOSI | Acc2: 87.9 -> 89.8 | Improved |
| add adaptive relaxed alignment (ablation) | CaReFlow · CMU-MOSI | Acc2: 88.5 -> 89.8 | Improved |
| add one-to-many mapping using pairs from different samples (ablation) | CaReFlow · CH-SIMS-v2 | Acc2: 79.8 -> 82.9 | Improved |
| replace simple MLP fusion with tensor fusion | CaReFlow · CH-SIMS-v2 | Acc2: 82.9 -> 83.6 | Improved |
| replace simple MLP fusion with low-rank modality fusion | CaReFlow · CH-SIMS-v2 | Acc5: 57.9 -> 58.0 | Improved |
| use CaReFlow instead of contrastive-learning alignment CLGSI | CaReFlow · CMU-MOSEI | Acc2: 87.2 -> 87.9 | Improved |
| vary number of Euler steps (ablation) | CaReFlow · CH-SIMS-v2 | Acc2 | No clear effect |
| use CaReFlow instead of AtCAF | CaReFlow · CMU-MOSEI | Acc7: 55.9 -> 55.7 | Hurt |
Closest papers by meaning
- Amplifying Prominent Representations in Multimodal Learning via Variational Dirichlet Process (NeurIPS 2025 · similarity 0.88)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.87)
- Strong Helps Weak: Directional Cross-Modal Alignment Transfer in Multi-modal LLMs (NeurIPS 2026 · similarity 0.87)
- Flowing from Words to Pixels: A Noise-Free Framework for Cross-Modality Evolution (CVPR 2025 · similarity 0.87)
- MSP: Modality Self-Play for Training Multimodal LLMs Without Paired Cross-Modal Data (NeurIPS 2026 · similarity 0.87)
- Beyond Modality Fusion: Deep Ensembles for Multimodal Classification (NeurIPS 2026 · similarity 0.87)
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.
