Research map / Vision-language and multimodal / Human and embodied modeling
MUSE: Resolving Manifold Misalignment in Visual Tokenization via Topological Orthogonality
Panqi Yang, Haodong Jing, Jiahao Chao, Tingyan Xiang, Li Lin, Yao Hu, Yang Luo, Yongqiang Ma
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 semantic anchoring (L_ITC) alone to reconstruction baseline (ablation) | reconstruction-only tokenizer · tokenizer ablation | rFID: 0.79 -> 1.88 | Hurt |
| add semantic anchoring (L_ITC) alone (ablation) | reconstruction-only tokenizer · tokenizer ablation | mIoU: 18.5 -> 14.1 | Hurt |
| add semantic anchoring (L_ITC) alone (ablation) | reconstruction-only tokenizer · tokenizer ablation | zero-shot accuracy: 12.4 -> 70.5 | Improved |
| add topology alignment (L_Topo) alone (ablation) | reconstruction-only tokenizer · tokenizer ablation | mIoU: 18.5 -> 32.5 | Improved |
| add topology alignment (L_Topo) alone (ablation) | reconstruction-only tokenizer · tokenizer ablation | MMBench: 35.6 -> 10.8 | Hurt |
| add L_Topo on top of L_ITC (full MUSE) (ablation) | MUSE with L_ITC only · tokenizer ablation | mIoU: 14.1 -> 46.5 | Improved |
| add L_Topo on top of L_ITC (full MUSE) (ablation) | MUSE with L_ITC only · tokenizer ablation | rFID: 1.88 -> 0.62 | Improved |
| add L_Topo on top of L_ITC (full MUSE) (ablation) | MUSE with L_ITC only · UMM downstream | MMBench: 65.2 -> 73.4 | Improved |
| use Synergistic Block decoupling instead of naive shared architecture (ablation) | MUSE · gradient conflict | MMBench: 65.2 -> 73.4 | Improved |
| use Synergistic Block instead of soft regularization (ablation) | MUSE · gradient conflict | mIoU: 22.1 -> 46.5 | Improved |
| use Synergistic Block instead of soft regularization (ablation) | MUSE · gradient conflict | gradient cosine similarity: -0.05 -> +0.04 | Improved |
| use full two-stream instead of Synergistic Block (ablation) | MUSE · gradient conflict | MMBench: 73.4 vs 73.6 | 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.
- UniLIP-3B — UniLiP: Adapting CLIP for Unified Multimodal Understanding, Generation and Editing (ICLR 2026) · better on 5 of 5 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
- Rethinking Visual Token Reduction in LVLMs Under Cross-Modal Misalignment (AAAI 2026 · similarity 0.88)
- Topology-Aware Representation Alignment for Semi-Supervised Vision-Language Learning (NeurIPS 2026 · similarity 0.88)
- TokenFlow: Unified Image Tokenizer for Multimodal Understanding and Generation (CVPR 2025 · similarity 0.88)
- VideoAnchor: Reinforcing Subspace-Structured Visual Cues for Coherent Visual-Spatial Reasoning (ICLR 2026 · similarity 0.87)
- End-to-End Vision Tokenizer Tuning (NeurIPS 2025 · similarity 0.87)
- UniFlow: A Unified Pixel Flow Tokenizer for Visual Understanding and Generation (ICLR 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.
