Towards Semantic Equivalence of Tokenization in Multimodal LLM
This addresses a key bottleneck in multimodal LLMs for researchers and practitioners by improving semantic alignment between vision and language, though it appears incremental as it builds on existing tokenization methods.
The paper tackles the problem of vision tokenization in multimodal LLMs, where existing methods fragment visual input and corrupt semantic integrity, by proposing a dynamic Semantic-Equivalent Vision Tokenizer (SeTok) that groups features into semantic units, resulting in superior performance across various tasks.
Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in processing vision-language tasks. One of the crux of MLLMs lies in vision tokenization, which involves efficiently transforming input visual signals into feature representations that are most beneficial for LLMs. However, existing vision tokenizers, essential for semantic alignment between vision and language, remain problematic. Existing methods aggressively fragment visual input, corrupting the visual semantic integrity. To address this, this paper proposes a novel dynamic Semantic-Equivalent Vision Tokenizer (SeTok), which groups visual features into semantic units via a dynamic clustering algorithm, flexibly determining the number of tokens based on image complexity. The resulting vision tokens effectively preserve semantic integrity and capture both low-frequency and high-frequency visual features. The proposed MLLM (Setokim) equipped with SeTok significantly demonstrates superior performance across various tasks, as evidenced by our experimental results. The project page is at https://chocowu.github.io/SeTok-web/.