Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Models

TL;DR

This paper systematically evaluates and analyzes the design choices of visually-conditioned language models (VLMs), providing standardized benchmarks, insights into their performance factors, and releasing improved models and tools.

Contribution

It introduces a comprehensive evaluation framework, investigates key design decisions, and releases new VLM checkpoints that outperform existing state-of-the-art models.

Findings

VLMs' performance varies significantly with design choices.

Pretrained visual representations and training strategies impact capabilities.

New VLMs at 7-13B scale outperform previous models.

Abstract

Visually-conditioned language models (VLMs) have seen growing adoption in applications such as visual dialogue, scene understanding, and robotic task planning; adoption that has fueled a wealth of new models such as LLaVa, InstructBLIP, and PaLI-3. Despite the volume of new releases, key design decisions around image preprocessing, architecture, and optimization are under-explored, making it challenging to understand what factors account for model performance $-$ a challenge further complicated by the lack of objective, consistent evaluations. To address these gaps, we first compile a suite of standardized evaluations spanning visual question answering, object localization, and challenge sets that probe properties such as hallucination; evaluations that provide fine-grained insight VLM capabilities. Second, we rigorously investigate VLMs along key design axes, including pretrained visual representations and training from base vs. instruct-tuned language models, amongst others. We couple our analysis with three resource contributions: (1) a unified framework for evaluating VLMs, (2) optimized, flexible training code, and (3) checkpoints for all models, including a family of VLMs at the 7-13B scale that strictly outperform InstructBLIP and LLaVa v1.5, the state-of-the-art in open VLMs.