Research map / Vision-language and multimodal / Cross-modal learning
CHARM: Collaborative Harmonization Across Arbitrary Modalities for Modality-Agnostic Semantic Segmentation
Lekang Wen, Jing Xiao, Liang Liao, Jiajun Chen, Mi 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 |
|---|---|---|---|
| add CoL (fusion via MPU-less collaborative learning) to addition fusion (ablation) | CHARM (MiT-B0) · DELIVER ablation | Average mIoU: 33.95 -> 41.53 | Improved |
| add MPU to CoL (ablation) | CHARM (MiT-B0) · DELIVER ablation | Last-1 mIoU: 3.28 -> 15.89 | Improved |
| add InE (with MPU removed variant d vs b) on top of CoL (ablation) | CHARM (MiT-B0) · DELIVER ablation | Last-1 mIoU: 3.28 -> 24.40 | Improved |
| add MPU to CoL+InE (ablation) | CHARM (MiT-B0) · DELIVER ablation | Average mIoU: 48.76 -> 50.97 | Improved |
| add MPU and CoL, Top-1 mIoU (ablation) | CHARM (MiT-B0) · DELIVER ablation | Top-1 mIoU: 63.17 -> 64.03 | No clear effect |
| add CoL without MPU to addition fusion (ablation) | CHARM (MiT-B0) · MUSES ablation | Last-1 mIoU: 2.26 -> 0.81 | Hurt |
| use softmin instead of softmax sampling in InE (ablation) | CHARM (MiT-B0) · DELIVER sampling ablation | Average mIoU: 46.80 -> 50.97 | Improved |
| use softmin instead of uniform sampling in InE (ablation) | CHARM (MiT-B0) · DELIVER sampling ablation | Last-1 mIoU: 23.36 -> 26.47 | Improved |
| use softmin instead of softmax sampling in InE (ablation) | CHARM (MiT-B0) · DELIVER sampling ablation | Top-1 mIoU: 64.03 -> 64.04 | No clear effect |
| replace MPU with explicit KL alignment (ablation) | CHARM · feature visualization | feature activations | Hurt |
| CHARM vs MAGIC on fragile-modal worst combination | CHARM (MiT-B2) · DELIVER | Last-1 mIoU: 0.27 -> 28.76 | Improved |
| CHARM vs MAGIC on MCubeS average (MiT-B0) | CHARM (MiT-B0) · MCubeS | Top-1 mIoU: 47.85 -> 48.24 | No clear effect |
Later papers that compared with this method
2 later accepted papers report Charm as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Probabilistic Prompt Adaptation for Unified Image Aesthetics and Quality Assessment (CVPR 2026 · better on 1 of 1 shared result)
- Fine-grained Image Aesthetic Assessment: Learning Discriminative Scores from Relative Ranks (CVPR 2026 · better on 0 of 1 shared result)
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.
- MAGIC — MAGIC: Multimodal relAtional Graph adversarIal inferenCe for Diverse and Unpaired Text-Based Image Captioning (AAAI 2022) · better on 10 of 10 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
- Keep the Balance: A Parameter-Efficient Symmetrical Framework for RGB+X Semantic Segmentation (CVPR 2025 · similarity 0.90)
- Linguistic-Aware Patch Slimming Framework for Fine-grained Cross-Modal Alignment (CVPR 2024 · similarity 0.87)
- DeepAlign: Mitigating Modality Conflict through Modality-Specific Alignment (CVPR 2026 · similarity 0.86)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.86)
- UniMRSeg: Unified Modality-Relax Segmentation via Hierarchical Self-Supervised Compensation (NeurIPS 2025 · similarity 0.86)
- OmniSegmentor: A Flexible Multi-Modal Learning Framework for Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
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.
