SOTAVerified

Improved Baselines with Visual Instruction Tuning

2023-10-05CVPR 2024Code Available6· sign in to hype

Haotian Liu, Chunyuan Li, Yuheng Li, Yong Jae Lee

arXivPDF

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this note, we show that the fully-connected vision-language cross-modal connector in LLaVA is surprisingly powerful and data-efficient. With simple modifications to LLaVA, namely, using CLIP-ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with simple response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~1 day on a single 8-A100 node. We hope this can make state-of-the-art LMM research more accessible. Code and model will be publicly available.

Tasks

Factual Inconsistency Detection in Chart CaptioningImage ClassificationReferring Expression ComprehensionReferring expression generationSpatial Reasoningvisual instruction followingVisual Question AnsweringVisual Question Answering (VQA)

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
CHOCOLATE-FTLLaVA-1.5-13BKendall's Tau-c0.21—Unverified
CHOCOLATE-LLMLLaVA-1.5-13BKendall's Tau-c0.06—Unverified
CHOCOLATE-LVLMLLaVA-1.5-13BKendall's Tau-c0—Unverified

Reproductions