Research map / Vision-language and multimodal / Multimodal LLMs
SPIDER: Multi-Layer Semantic Token Pruning and Adaptive Sub-Layer Skipping in Multimodal Large Language Models
Tianxiang Chen, Zhentao Tan, Zi Ye, Yue Wu, Xiaobing Tu, Jinkui Ren, Xiantao Zhang, Tao Gong, Qi Chu, Nenghai Yu, Xipeng Qiu, Jieping Ye
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 clustering (ablation) | MSV-Prune · LLaVA-NeXT-7B 1920 tokens | relative accuracy: 99.33% -> 100.64% | Improved |
| add middle-layer tokens (ablation) | MSV-Prune · LLaVA-NeXT-7B 1920 tokens | relative accuracy: 99.13% -> 100.64% | Improved |
| add inter-layer similarity term (ablation) | MSV-Prune · LLaVA-NeXT-7B 1920 tokens | relative accuracy: 99.60% -> 100.64% | Improved |
| add similarity with anchor tokens (ablation) | MSV-Prune · LLaVA-NeXT-7B 1920 tokens | relative accuracy: 100.20% -> 100.64% | Improved |
| use 4 clusters instead of 8 (ablation) | MSV-Prune · LLaVA-NeXT-7B 1920 tokens | relative accuracy: 99.68% -> 100.64% | Improved |
| use anchor ratio 0.7 instead of 0 (ablation) | MSV-Prune · LLaVA-1.5-7B | relative accuracy: 95.74% -> 99.56% | Improved |
| use anchor ratio 0.7 instead of 1.0 (ablation) | MSV-Prune · LLaVA-1.5-7B | relative accuracy: 98.74% -> 99.56% | Improved |
| add sub-layer contribution (SLC) to fusion score (ablation) | ASL-Skip · LLaVA-1.5-7B | relative accuracy: 98.30% -> 99.81% | Improved |
| add online skippability score S_sa (ablation) | ASL-Skip · LLaVA-1.5-7B | relative accuracy: 99.17% -> 99.81% | Improved |
| add accumulation of skip score (ablation) | ASL-Skip · LLaVA-1.5-7B | relative accuracy: 99.22% -> 99.81% | Improved |
| skip all FFN sub-layers for visual tokens (ablation) | ASL-Skip · LLaVA-1.5-7B | relative accuracy: 99.81% -> 71.34% | Hurt |
| fine-tune SPIDER (training-aware) instead of training-free | SPIDER · LLaVA-1.5-7B ~30% TFLOPs | relative accuracy: 99.0% -> 99.5% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VisionZip — VisionZip: Longer is Better but Not Necessary in Vision Language Models (CVPR 2025) · better on 2 of 2 shared results
- SparseVLM — SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference (ICML 2025) · better on 2 of 2 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
- ATP-LLaVA: Adaptive Token Pruning for Large Vision Language Models (CVPR 2025 · similarity 0.93)
- Hi-Lo Prune: Look at What You'll Lose before Pruning with Hierarchical Token Selection (CVPR 2026 · similarity 0.92)
- One Layer's Trash is Another Layer's Treasure: Adaptive Layer-wise Visual Token Selection in LVLMs (CVPR 2026 · similarity 0.91)
- When Token Pruning is Worse than Random: Understanding Visual Token Information in VLLMs (CVPR 2026 · similarity 0.91)
- What Do Visual Tokens Really Encode? Uncovering Sparsity and Redundancy in Multimodal Large Language Models (CVPR 2026 · similarity 0.91)
- Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization (NeurIPS 2025 · similarity 0.91)
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.
