Research map / Vision-language and multimodal / Multimodal LLMs
FlexAttention for Efficient High-Resolution Vision-Language Models
Junyan Li, Delin Chen, Tianle Cai, Peihao Chen, Yining Hong, Zhenfang Chen, Yikang Shen, Chuang Gan
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 FlexAttention high-resolution input (1008x1008) to LLaVA-1.5-7b | LLaVA-1.5-7b · V* Bench | overall accuracy: 47.6 -> 54.5 | Improved |
| add FlexAttention to LLaVA-1.5-7b | LLaVA-1.5-7b · MagnifierBench | accuracy: 26.8 -> 35.0 | Improved |
| FlexAttention instead of LLaVA-HD or LLaVA-XAttn high-resolution handling | LLaVA-1.5-7b · MagnifierBench | TFLOPs: 24.9 -> 17.1 (vs LLaVA-HD); 27.1 -> 17.1 (vs LLaVA-XAttn) | Improved |
| select high-resolution tokens by attention map instead of random or center selection (ablation) | FlexAttention · MagnifierBench and TextVQA | accuracy | Improved |
| increase high-resolution input from 672 to 1008 (ablation) | FlexAttention · TextVQA | accuracy | Improved |
| increase high-resolution input from 1008 to 1344 (ablation) | FlexAttention · TextVQA | accuracy | No clear effect |
| increase high-resolution input resolution (ablation) | FlexAttention · MagnifierBench | accuracy | Improved |
| add FlexAttention to LLaVA-1.5-7b | LLaVA-1.5-7b · RefCOCO val, small objects | accuracy: 41.3 -> 51.3 | Improved |
| add FlexAttention to LLaVA-1.5-7b | LLaVA-1.5-7b · MME and MM-Vet | score: MME 1511 -> 1479; MM-Vet 31.1 -> 29.4 | Hurt |
| add FlexAttention to LLaVA-1.5-7b | LLaVA-1.5-7b · POPE, GQA, VQAv2 | accuracy: POPE 85.9 -> 85.9 | 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.
- GeoChat — GeoChat: Grounded Large Vision-Language Model for Remote Sensing (CVPR 2024) · 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
- Hi-Lo Prune: Look at What You'll Lose before Pruning with Hierarchical Token Selection (CVPR 2026 · similarity 0.88)
- ATP-LLaVA: Adaptive Token Pruning for Large Vision Language Models (CVPR 2025 · similarity 0.88)
- LearnPruner: Rethinking Attention-based Token Pruning in Vision Language Models (ICLR 2026 · similarity 0.87)
- TransPrune: Token Transition Pruning for Efficient Large Vision-Language Model (CVPR 2026 · similarity 0.87)
- Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization (NeurIPS 2025 · similarity 0.87)
- SCoRe: Salience-Coverage Reduction for Vision Token Pruning in Vision-Language Models (CVPR 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.
