Research map / Vision-language and multimodal / Multimodal LLMs
Fourier Compressor: Frequency-Domain Visual Token Compression for Vision-Language Models
Huanyu Wang, Jushi Kai, Haoli Bai, Lu Hou, Bo Jiang, Ziwei He, Zhouhan Lin
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 |
|---|---|---|---|
| apply Fourier Compressor to 256 tokens (55.6% compression) | LLaVA-v1.5-7B · Image benchmarks avg | average score: 64.6 -> 64.6 | No clear effect |
| apply Fourier Compressor to 144 tokens (75% compression) | LLaVA-v1.5-7B · Image benchmarks avg | average score: 64.6 -> 64.6 | No clear effect |
| compress to 36 tokens (93.75%) | LLaVA-v1.5-7B · Image benchmarks avg | average score: 64.6 -> 62.1 | Hurt |
| use Fourier Compressor instead of MQT-LLaVA | LLaVA-v1.5-7B · Image benchmarks avg (144 tokens) | average score: 62.8 -> 64.6 | Improved |
| use Fourier Compressor instead of MQT-LLaVA | LLaVA-v1.5-7B · Image benchmarks avg (36 tokens) | average score: 61.1 -> 62.1 | Improved |
| apply Fourier Compressor to 144 tokens | LLaVA-v1.5-13B · Image benchmarks avg | average score: 67.3 -> 65.9 | Hurt |
| use Fourier Compressor instead of PruMerge | LLaVA-v1.5-13B · Image benchmarks (144 tokens) | POPE: 84.4 -> 85.4 | Improved |
| apply Fourier Compressor (57.3% fewer tokens) | Qwen2-VL-2B · Image benchmarks avg | average score: 66.8 -> 65.7 | Hurt |
| apply Fourier Compressor (57.3% fewer tokens) | Qwen2.5-VL-3B · Image benchmarks avg | average score: 71.1 -> 72.6 | Improved |
| apply Fourier Compressor (57.3% fewer tokens) | Qwen2.5-VL-3B · RealWorldQA | accuracy: 59.6 -> 64.3 | Improved |
| apply Fourier Compressor to 36 tokens | LLaVA-v1.5-7B · FLOPs | FLOPs (T): 8.54 -> 1.38 | Improved |
| apply Fourier Compressor (87.5% fewer video tokens) | LLaVA-v1.5-7B · Video benchmark avg (zero-shot) | average score: 45.6 -> 44.0 | Hurt |
Closest papers by meaning
- ApET: Approximation-Error Guided Token Compression for Efficient VLMs (CVPR 2026 · similarity 0.91)
- Hybrid Token Compression for Vision-Language Models (CVPR 2026 · similarity 0.91)
- PVC: Progressive Visual Token Compression for Unified Image and Video Processing in Large Vision-Language Models (CVPR 2025 · similarity 0.91)
- LLMC+: Benchmarking Vision-Language Model Compression with a plug-and-play Toolkit (AAAI 2026 · similarity 0.91)
- Prune Redundancy, Preserve Essence: Vision Token Compression in VLMs via Synergistic Importance-Diversity (ICLR 2026 · similarity 0.91)
- DUET-VLM: Dual stage Unified Efficient Token reduction for VLM Training and Inference (CVPR 2026 · 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.
