Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Shaking Up VLMs: Comparing Transformers and Structured State Space Models for Vision & Language Modeling

About

This study explores replacing Transformers in Visual Language Models (VLMs) with Mamba, a recent structured state space model (SSM) that demonstrates promising performance in sequence modeling. We test models up to 3B parameters under controlled conditions, showing that Mamba-based VLMs outperforms Transformers-based VLMs in captioning, question answering, and reading comprehension. However, we find that Transformers achieve greater performance in visual grounding and the performance gap widens with scale. We explore two hypotheses to explain this phenomenon: 1) the effect of task-agnostic visual encoding on the updates of the hidden states, and 2) the difficulty in performing visual grounding from the perspective of in-context multimodal retrieval. Our results indicate that a task-aware encoding yields minimal performance gains on grounding, however, Transformers significantly outperform Mamba at in-context multimodal retrieval. Overall, Mamba shows promising performance on tasks where the correct output relies on a summary of the image but struggles when retrieval of explicit information from the context is required.

Georgios Pantazopoulos, Malvina Nikandrou, Alessandro Suglia, Oliver Lemon, Arash Eshghi• 2024

Related benchmarks

TaskDatasetResultRank
Visual Question AnsweringTextVQA (val)
VQA Score42.1
365
Visual Question AnsweringGQA (test-dev)
Accuracy60.4
236
Visual Question AnsweringVQA 2.0 (val)
Accuracy (Overall)76.1
183
Object Hallucination EvaluationPOPE (test)
Accuracy87.3
107
Visual Question AnsweringAI2D (test)
Accuracy83.7
82
Showing 5 of 5 rows

Other info

Follow for update