UniBench: Visual Reasoning Requires Rethinking Vision-Language Beyond Scaling

TL;DR

UniBench provides a comprehensive, unified framework to evaluate vision-language models across diverse capabilities, revealing that scaling data or models alone does not enhance reasoning skills and highlighting the importance of data quality and tailored objectives.

Contribution

The paper introduces UniBench, a unified benchmarking suite for over 50 vision-language tasks, enabling systematic evaluation and comparison of models at scale.

Findings

Scaling data or model size improves many capabilities but not reasoning.

Current top models struggle with simple digit recognition tasks.

Data quality and tailored training objectives are more effective than scaling alone.

Abstract

Significant research efforts have been made to scale and improve vision-language model (VLM) training approaches. Yet, with an ever-growing number of benchmarks, researchers are tasked with the heavy burden of implementing each protocol, bearing a non-trivial computational cost, and making sense of how all these benchmarks translate into meaningful axes of progress. To facilitate a systematic evaluation of VLM progress, we introduce UniBench: a unified implementation of 50+ VLM benchmarks spanning a comprehensive range of carefully categorized capabilities from object recognition to spatial awareness, counting, and much more. We showcase the utility of UniBench for measuring progress by evaluating nearly 60 publicly available vision-language models, trained on scales of up to 12.8B samples. We find that while scaling training data or model size can boost many vision-language model capabilities, scaling offers little benefit for reasoning or relations. Surprisingly, we also discover today's best VLMs struggle on simple digit recognition and counting tasks, e.g. MNIST, which much simpler networks can solve. Where scale falls short, we find that more precise interventions, such as data quality or tailored-learning objectives offer more promise. For practitioners, we also offer guidance on selecting a suitable VLM for a given application. Finally, we release an easy-to-run UniBench code-base with the full set of 50+ benchmarks and comparisons across 59 models as well as a distilled, representative set of benchmarks that runs in 5 minutes on a single GPU.