AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

UltraViT: Latency-Optimized On-device Vision Encoder for Large Vision-Language Models

Ioannis Maniadis Metaxas, Adrian Bulat, Alberto Baldrati, Anestis Zaganidis, Yassine Ouali, Hyeonuk Kim, Georgios Tzimiropoulos

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 0 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
use heterogeneous mixers (C, C+S, CP, A) instead of all-convolutional (ablation)UltraViT · LLaVA-OV 1.5B avg of 7 benchmarksavg score: 56.4 -> 58.5Improved
channel-partitioned attention in stage 3 instead of full single-head attention (ablation)UltraViT · LLaVA-OV 1.5B avgavg score / throughput: 58.5 vs 58.5; 148.9 vs 126.7 inf/sNo clear effect
dense distillation instead of contrastive training (ablation)UltraViT · DocVQAaccuracy: 58.9 -> 63.9Improved
dense distillation instead of contrastive training (ablation)FastViT · DocVQAaccuracy: 60.7 -> 62.1Improved
use larger teacher for dense distillation (ablation)UltraViT · DocVQAaccuracy: 63.9 -> 67.5Improved
add generative pre-training with frozen 1.5B LLM (ablation)UltraViT · DocVQAaccuracy: 67.5 -> 72.0Improved
generative pre-training with 0.5B instead of 1.5B LLM (ablation)UltraViT · DocVQAaccuracy: 72.0 -> 70.0Hurt
dynamic 0.5B/1.5B LLM switching vs pure 1.5B (ablation)UltraViT · InfoVQAaccuracy: 46.8 -> 47.9No clear effect
choice of teacher for dense pretraining (ablation)UltraViT · DocVQAaccuracy: 63.9 (SigLIP2-NF) vs 65.3 (SigLIP2)Mixed
UltraVLM vs re-trained FastVLMUltraViT · TextVQAaccuracy: 62.9 -> 68.9Improved
UltraVLM vs re-trained FastVLMUltraViT · MMMUaccuracy: 41.1 -> 38.1Hurt
UltraVLM vs token reduction methodsUltraViT · DocVQAaccuracy: 55.1 (VisionZip) -> 71.8Improved

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