AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Oasis: One Image is All You Need for Multimodal Instruction Data Synthesis

Letian Zhang, Quan Cui, Bingchen Zhao, Cheng Yang

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Visual instruction tuning · 17 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 Oasis synthetic data to LLaVA-NeXT SFT dataLLaVA-NeXT Vicuna-7B-v1.5 · 14-benchmark averageaverage score: 53.0 -> 56.1Improved
add Oasis synthetic data to LLaVA-NeXT SFT dataLLaVA-NeXT Qwen2.5-7B · 14-benchmark averageaverage score: 61.4 -> 63.2Improved
add Oasis synthetic data to LLaVA-NeXT SFT dataLLaVA-NeXT Llama3-8B · 14-benchmark averageaverage score: 55.8 -> 59.0Improved
add Oasis data in place of Cambrian original annotationsVicuna-7B-v1.5 · 14-benchmark averageaverage score: 54.9 -> 56.1Improved
add Oasis data instead of MMEvol data of equal sizeVicuna-7B-v1.5 · 14-benchmark averageaverage score: 51.6 -> 56.1Improved
scale Oasis data from 0 to 150k added to 100k SFTLLaVA-100k · 14-benchmark averageaverage score: 46.5 -> 46.6No clear effect
scale Oasis data from 0 to 500k added to 100k SFTLLaVA-100k · 14-benchmark averageaverage score: 46.5 -> 51.7Improved
add Oasis data to baselineVicuna-7B-v1.5 · DocVQAaccuracy: 71.7 -> 76.0Improved
add 70k OCR-domain Oasis dataLLaVA-NeXT Vicuna-7B-v1.5 · DocVQAaccuracy: 71.7 -> 75.0Improved
add 2k medical Oasis data to 2k LLaVA dataLLaVA-NeXT · PathVQAaccuracy: 34.0 -> 41.2Improved
add 250k recycled caption data (ablation)Oasis · 14-benchmark averageaverage score: 56.1 -> 56.4Improved
add instruction quality control (ablation)Oasis (200k) · 14-benchmark averageaverage score: 53.9 -> 54.9Improved

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