AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

The Scalability of Simplicity: Empirical Analysis of Vision-Language Learning with a Single Transformer

Weixian Lei, Jiacong Wang, Haochen Wang, Xiangtai Li, Jun Hao Liew, Jiashi Feng, Zilong Huang

ICCV 2025 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment · 26 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 bidirectional attention among image patches with multimodal RoPE (ablation)SAIL-0.5B · TextVQAaccuracy: 16.2 -> 20.1Improved
add bidirectional attention among image patches with multimodal RoPE (ablation)SAIL-0.5B · VQAv2accuracy: 57.8 -> 59.1Improved
add bidirectional attention among image patches with multimodal RoPE (ablation)SAIL-0.5B · GQAaccuracy: 45.2 -> 46.9Improved
add bidirectional attention among image patches with multimodal RoPE (ablation)SAIL-0.5B · SEEDBench-IMGaccuracy: 33.8 -> 35.1Improved
interleave pure text data (SlimPajama) in pretraining (ablation)SAIL-0.5B · ScienceQAaccuracy: 48.6 -> 59.6Improved
interleave pure text data (SlimPajama) in pretraining (ablation)SAIL-0.5B · GQAaccuracy: 42.1 -> 46.9Improved
interleave pure text data (SlimPajama) in pretraining (ablation)SAIL-0.5B · TextVQAaccuracy: 18.3 -> 20.1Improved
scale pretraining data from 32M to 512M pairs (single transformer vs modular MLLM)SAIL · Normalized average benchmark performanceperformanceImproved
scale model size from 0.5B to 3B to 7BSAIL · downstream VLM tasks averageaverage performanceImproved
single transformer (SAIL) vs modular MLLM on spatial reasoningSAIL · MMBench Physical Relationaccuracy: 30.4 -> 52.2Improved
single transformer (SAIL) vs modular MLLM on spatial positionSAIL · MME Positionscore: 98.3 -> 160.0Improved
single transformer (SAIL) vs modular MLLM on celebrity knowledgeSAIL · MME Celebrityscore: 134.0 -> 108.2Hurt

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