When Big Models Train Small Ones: Label-Free Model Parity Alignment for Efficient Visual Question Answering using Small VLMs
This work addresses the problem of inefficient large models for resource-constrained settings in VQA, offering an incremental improvement through targeted knowledge transfer.
The paper tackles the performance gap between large and small vision-language models (VLMs) in visual question answering (VQA) by introducing the Model Parity Aligner (MPA), a framework that uses unlabeled images and knowledge transfer to improve small VLMs, achieving consistent performance gains across four diverse VQA benchmarks.
Large Vision-Language Models (L-VLMs) have demonstrated remarkable performance in various vision and language tasks, including visual question answering (VQA). However, their high computational cost makes them impractical for resource-constrained settings and inference-heavy applications. In contrast, Small Vision-Language Models (S-VLMs) offer efficiency but suffer from a significant performance gap compared to their larger counterparts. In this work, we introduce the Model Parity Aligner (MPA), a novel framework designed to systematically improve S-VLMs by leveraging unlabeled images and effective knowledge transfer from L-VLMs. Instead of traditional knowledge distillation methods that rely on labeled training data, MPA employs a strategic parity-based approach that precisely identifies the knowledge disparities between S-VLMs and L-VLMs, and optimizes training by targeting only these disparities. We conduct extensive experiments on four diverse VQA benchmarks, namely TextVQA, ST-VQA, ChartQA, and OKVQA, each of which requires specialized reasoning capabilities such as text recognition, chart interpretation, and commonsense and factual understanding. Our results demonstrate that MPA consistently enhances the performance of S-VLMs on all benchmarks, reducing the performance gap while maintaining computational efficiency. We make our code publicly available.