Research map / Vision-language and multimodal / Multimodal LLMs
SPAR: Semantic-Pixel Self-Alignment and Adaptive Routing for Unified Multimodal Models
Hongxiang Li, Hongxu chen, chenyang zhu, Xiaoshuang Huang, Jiayin Cai, Xiaolong Jiang, Yao Hu, Long Chen
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 pixel stream to the tokenizer (ablation) | SPAR · ImageNet 50k reconstruction | PSNR: 24.62 -> 26.65 | Improved |
| add pixel stream to the tokenizer (ablation) | SPAR · ImageNet 50k reconstruction | SSIM: 0.788 -> 0.856 | Improved |
| add pixel stream to the tokenizer (ablation) | SPAR · MMBench | accuracy: 72.6 -> 73.0 | No clear effect |
| add semantic stream to the tokenizer (ablation) | SPAR · MMBench | accuracy: 18.4 -> 73.0 | Improved |
| add semantic stream to the tokenizer (ablation) | SPAR · MME-P | score: 709 -> 1500 | Improved |
| add semantic stream to the tokenizer (ablation) | SPAR · ImageNet 50k reconstruction | PSNR: 25.28 -> 26.65 | Improved |
| unfreeze encoder with dual-stream training instead of freezing (ablation) | SPAR · ImageNet 50k reconstruction | rFID: 6.14 -> 0.27 | Improved |
| unfreeze encoder with dual-stream training instead of freezing (ablation) | SPAR · MMBench | accuracy: 72.6 -> 73.0 | Improved |
| add Dynamic Token Routing instead of last-layer extraction (ablation) | SPAR · GenEval | overall: 0.86 -> 0.89 | Improved |
| add Dynamic Token Routing instead of last-layer extraction (ablation) | SPAR · ImgEdit | overall: 3.73 -> 3.85 | Improved |
| add self-alignment loss (ablation) | SPAR · GenEval | overall: 0.86 -> 0.89 | Improved |
| add self-alignment loss (ablation) | SPAR · WISE | overall: 0.53 -> 0.57 | 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.
- OmniGen2 — OmniGen2: Towards Instruction-Aligned Multimodal Generation (CVPR 2026) · better on 1 of 1 shared result
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
- Vision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations (NeurIPS 2025 · similarity 0.91)
- UVU: Improving Multimodal Understanding via Vision-Language Unified Autoregressive Paradigm (CVPR 2026 · similarity 0.90)
- Analyzing Fine-Grained Alignment and Enhancing Vision Understanding in Multimodal Language Models (NeurIPS 2025 · similarity 0.90)
- Reconstruction Alignment Improves Unified Multimodal Models (ICLR 2026 · similarity 0.90)
- Seeing What Matters: A Training-Free Self-Guided Framework for Multimodal Detail Perception and Reasoning (CVPR 2026 · similarity 0.90)
- Seeing is Understanding: Unlocking Causal Attention into Modality-Mutual Attention for Multimodal LLMs (ICML 2026 · similarity 0.90)
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.
