In-Context Learning Improves Compositional Understanding of Vision-Language Models
This work addresses a specific bottleneck in VLMs for researchers and practitioners, but it appears incremental as it builds on existing ICL methods applied to a known issue.
The paper tackled the problem of compositional image understanding in Vision-Language Models (VLMs) by benchmarking their capabilities and proposing In-Context Learning (ICL) to enhance reasoning, resulting in outperformance over baseline models on multiple datasets.
Vision-Language Models (VLMs) have shown remarkable capabilities in a large number of downstream tasks. Nonetheless, compositional image understanding remains a rather difficult task due to the object bias present in training data. In this work, we investigate the reasons for such a lack of capability by performing an extensive bench-marking of compositional understanding in VLMs. We compare contrastive models with generative ones and analyze their differences in architecture, pre-training data, and training tasks and losses. Furthermore, we leverage In-Context Learning (ICL) as a way to improve the ability of VLMs to perform more complex reasoning and understanding given an image. Our extensive experiments demonstrate that our proposed approach outperforms baseline models across multiple compositional understanding datasets.