VDInstruct: Zero-Shot Key Information Extraction via Content-Aware Vision Tokenization
This addresses inefficiencies in document understanding for applications like processing receipts and contracts, offering a novel method that is incremental but with strong specific gains.
The paper tackled the problem of inefficient vision tokenization in multimodal large language models for key information extraction from visual documents, achieving state-of-the-art results with a 3.6x reduction in image tokens and a +5.5 F1 point improvement in zero-shot evaluations.
Key Information Extraction (KIE) underpins the understanding of visual documents (e.g., receipts and contracts) by extracting precise semantic content and accurately capturing spatial structure. Yet existing multimodal large language models (MLLMs) often perform poorly on dense documents and rely on vision tokenization approaches that scale with image size, leading to redundant computation and memory inefficiency. To address these challenges, we introduce VDInstruct, an MLLM that separates spatial region detection from semantic feature extraction. Central to our model is a content-aware tokenization strategy: rather than fragmenting the entire image uniformly, it generates tokens in proportion to document complexity, preserving critical structure while eliminating wasted tokens. Leveraging a three-stage training paradigm, our model achieves state-of-the-art (SOTA) results on KIE benchmarks, matching or exceeding the accuracy of leading approaches while reducing the number of image tokens by roughly 3.6x. In zero-shot evaluations, VDInstruct surpasses strong baselines-such as DocOwl 1.5-by +5.5 F1 points, highlighting its robustness to unseen documents. These findings show that content-aware tokenization combined with explicit layout modeling offers a promising direction forward for document understanding. Data, source code, and model weights will be made publicly available.