AI Paper Atlas

Research map / Video / Video-language understanding

ParallelVLM: Lossless Video-LLM Acceleration with Visual Alignment Aware Parallel Speculative Decoding

Quan Kong, Yuhao Shen, Yicheng Ji, Huan Li, Cong Wang

CVPR 2026 · Video › Video-language understanding · approach: Long video understanding · 9 citations (OpenAlex)

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 changedWhereMeasuredEffect
raise draft pruning ratio from 0.0 to 0.9 (ablation)ParallelVLM · LLaVA-OV 7B/72Bspeedup: 2.94x -> 3.36xImproved
raise draft pruning ratio from 0.0 to 0.9 (ablation)ParallelVLM · Qwen2.5-VL 7B/32Bspeedup: 2.08x -> 2.42xImproved
raise draft pruning ratio from 0.0 to 0.9 (ablation)draft model speed ratio · LLaVA-OV 7B/72Bspeed ratio c: 5 -> 9Improved
push pruning ratio to 1.0 (prune all draft visual tokens) (ablation)ParallelVLM · LLaVA-OV 7B/72B VideoDetailCaptionspeedup: 3.43x -> 2.53xHurt
push pruning ratio to 0.99 (ablation)ParallelVLM · LLaVA-OV 7B/72B VideoDetailCaptionspeedup: 3.43x -> 2.90xHurt
use UV-Prune verifier-guided pruning instead of SpecVLM attention-guided pruning (ablation)ParallelVLM · pruning alignment, alpha=0.9accepted length M: 0.2 drop for SpecVLMImproved
add FlashAttention (ablation)ParallelVLM · LLaVA-OV-72B, 8K sequencespeedup: 1.22xImproved
add FlashAttention (ablation)ParallelVLM · LLaVA-OV-72B, 32K sequencespeedup: 1.22x -> 1.41xImproved
use ParallelVLM instead of SpecVLMlossless SD · LLaVA-OV 7B/72Bavg speedup: 2.74x -> 3.36xImproved
use ParallelVLM instead of OnSDlossless SD · LLaVA-OV 7B/7Bavg speedup: 0.84x -> 1.55xImproved
use ParallelVLM instead of FastV token pruningLLaVA-OV-72B · 10% retentionspeedup: 1.63x -> 3.36xImproved
use ParallelVLM instead of FastV token pruningLLaVA-OV-72B · 10% retentiontoken acceptance ratio: 84.6% -> 99.1%Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy