DeepStack: Deeply Stacking Visual Tokens is Surprisingly Simple and Effective for LMMs
This addresses efficiency and performance bottlenecks in multimodal AI for researchers and practitioners, offering a simple yet effective architectural change.
The paper tackles the high computational and memory costs in large multimodal models (LMMs) by introducing DeepStack, a new architecture that stacks visual tokens across transformer layers, resulting in improved performance with minimal additional cost. For example, DeepStack 7B and 13B models achieve average improvements of 2.7 and 2.9 points across 9 benchmarks, with gains up to 11.0 points on specific high-resolution tasks.
Most large multimodal models (LMMs) are implemented by feeding visual tokens as a sequence into the first layer of a large language model (LLM). The resulting architecture is simple but significantly increases computation and memory costs, as it has to handle a large number of additional tokens in its input layer. This paper presents a new architecture DeepStack for LMMs. Considering $N$ layers in the language and vision transformer of LMMs, we stack the visual tokens into $N$ groups and feed each group to its aligned transformer layer \textit{from bottom to top}. Surprisingly, this simple method greatly enhances the power of LMMs to model interactions among visual tokens across layers but with minimal additional cost. We apply DeepStack to both language and vision transformer in LMMs, and validate the effectiveness of DeepStack LMMs with extensive empirical results. Using the same context length, our DeepStack 7B and 13B parameters surpass their counterparts by \textbf{2.7} and \textbf{2.9} on average across \textbf{9} benchmarks, respectively. Using only one-fifth of the context length, DeepStack rivals closely to the counterparts that use the full context length. These gains are particularly pronounced on high-resolution tasks, e.g., \textbf{4.2}, \textbf{11.0}, and \textbf{4.0} improvements on TextVQA, DocVQA, and InfoVQA compared to LLaVA-1.5-7B, respectively. We further apply DeepStack to vision transformer layers, which brings us a similar amount of improvements, \textbf{3.8} on average compared with LLaVA-1.5-7B.