PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
October 2, 20250 citationsOpen Access

VDInstruct: Zero-Shot Key Information Extraction via Content-Aware Vision Tokenization

View Full Paper
SNSon NguyenGNGiang NguyenHDHung Dao

Key Points

  • VDInstruct achieves state-of-the-art results in key information extraction benchmarks by effectively tokenizing complex documents.
  • The new model reduces the number of image tokens by approximately 3.6 times compared to previous methods, enhancing computational efficiency.
  • In zero-shot evaluations, VDInstruct outperforms DocOwl 1.5 by 5.5 F1 points, highlighting its robustness with unseen documents.
  • This approach integrates content-aware tokenization with layout modeling to advance document understanding techniques.

Abstract

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.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Nguyen et al. (2025) studied this question.

synapsesocial.com/papers/68de5da283cbc991d0a20907https://doi.org/10.48550/arxiv.2507.09531
Ask AI
Helpful
Bookmark
Share
View Full Paper