AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Let ViT Speak: Generative Language-Image Pre-training

Yan Fang, Mengcheng Lan, Zilong Huang, Weixian Lei, Yunqing Zhao, Yujie Zhong, Yingchen Yu, Qi She, Yao Zhao, Yunchao Wei

ECCV 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Vision-language pretraining · 2 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
add gated attention instead of 4 register tokens (ablation)GenLIP-So/16 · Attention-sink ablation, GenLIP-So/16, 1B samplesALL AVG: 55.1 -> 56.3Improved
add gated attention instead of 1 register token (ablation)GenLIP-So/16 · Attention-sink ablation, GenLIP-So/16, 1B samplesALL AVG: 53.3 -> 56.3Improved
add gated attention to Qwen3-initialized SAIL (ablation)SAIL · SAIL comparison, Qwen3-0.6B arch, 1B samplesALL AVG: 53.6 -> 54.8Improved
train SAIL-g from scratch instead of Qwen3 initialization (ablation)SAIL-g · SAIL comparison, Qwen3-0.6B arch, 1B samplesALL AVG: 54.8 -> 56.0Improved
add diverse-resolution adaptation stage 2 (L/16) (ablation)GenLIP · Frozen encoder, LLaVA-NeXT Qwen2.5-1.5BALL AVG: 55.2 -> 61.5Improved
add diverse-resolution adaptation stage 2 (So/16) (ablation)GenLIP · Frozen encoder, LLaVA-NeXT Qwen2.5-1.5BALL AVG: 58.9 -> 62.6Improved
add diverse-resolution adaptation stage 2 (g/16) (ablation)GenLIP · Frozen encoder, LLaVA-NeXT Qwen2.5-1.5BALL AVG: 60.0 -> 65.2Improved
native-aspect-ratio adaptation stage 2 (ablation)GenLIP-g/16 · ImageNet-1K frozen attentive probetop-1 accuracy: 85.2 -> 83.0Hurt
scale model from L/16 to g/16 (ablation)GenLIP · Frozen encoder, LLaVA-NeXT Qwen2.5-1.5BALL AVG: 61.5 -> 65.2Improved
GenLIP vs SigLIP at equal 2.0B dataGenLIP-So/16 · Frozen encoder, same 2.0B samples, Qwen2.5-1.5BALL AVG: 54.4 -> 57.2Improved
GenLIP vs OpenVision2 on OCRBench at equal dataGenLIP-So/16 · Frozen encoder, same 2.0B samples, Qwen2.5-1.5BOCRBench: trails by 6.3Hurt
GenLIP vs SigLIP2 (40B pairs) with 7B LLMGenLIP · Frozen encoder, LLaVA-NeXT Qwen2.5-7BALL AVG: 69.4 -> 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