Research map / Vision-language and multimodal / Cross-modal learning
CoRiM: Conflict-driven Risk Minimization for Dynamic Multimodal Fusion
Shihao Zou, Wei Wei
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 all three risk terms instead of fused uncertainty only (ablation) | CoRiM · MVSA (eps=5, avg) | accuracy: 71.87 -> 75.65 | Improved |
| use all three risk terms instead of JS consistency only (ablation) | CoRiM · MVSA (eps=10, avg) | accuracy: 60.50 -> 65.34 | Improved |
| use JS divergence instead of symmetric KL (ablation) | CoRiM · MVSA (eps=10, avg) | accuracy: 63.01 -> 65.34 | Improved |
| use JS divergence instead of Hellinger (ablation) | CoRiM · MVSA (eps=5, avg) | accuracy: 72.29 -> 75.65 | Improved |
| use per-sample weights instead of global weights (ablation) | CoRiM · MVSA (eps=5, avg) | accuracy: 68.54 -> 75.65 | Improved |
| use per-sample weights instead of class-level weights (ablation) | CoRiM · MVSA (eps=10, avg) | accuracy: 64.05 -> 65.34 | Improved |
| use Frank-Wolfe instead of projected gradient descent (ablation) | CoRiM · MVSA (eps=10, avg) | accuracy: 59.99 -> 65.34 | Improved |
| use Frank-Wolfe instead of entropic mirror descent (ablation) | CoRiM · MVSA (eps=10, avg) | accuracy: 61.40 -> 65.34 | Improved |
| replace scalar-proxy weighting (PDF) with conflict risk minimization | multimodal fusion · MVSA (eps=10, avg) | accuracy: 63.09 -> 65.34 | Improved |
| replace ReconBoost with CoRiM on three-modality data | multimodal fusion · CH-SIMS | accuracy: 73.88 -> 73.68 | No clear effect |
| replace ReconBoost with CoRiM on three-modality data | multimodal fusion · MOSEI | accuracy: 68.61 -> 69.21 | Improved |
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.
- ReconBoost — ReconBoost: Boosting Can Achieve Modality Reconcilement (ICML 2024) · better on 2 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
- Predictive Dynamic Fusion (ICML 2024 · similarity 0.88)
- Test-Time Dynamic Image Fusion (NeurIPS 2024 · similarity 0.84)
- Coupled Mamba: Enhanced Multimodal Fusion with Coupled State Space Model (NeurIPS 2024 · similarity 0.83)
- Towards Robust Multimodal Open-set Test-time Adaptation via Adaptive Entropy-aware Optimization (ICLR 2025 · similarity 0.83)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.83)
- Conditional Controllable Image Fusion (NeurIPS 2024 · similarity 0.83)
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.
