Papers by Yiheng Xu

8 papers
XFUND: A Benchmark Dataset for Multilingual Visually Rich Form Understanding (2022.findings-acl)

Copied to clipboard

Challenge: Existing research on multimodal pre-training for visually rich document understanding tasks has focused on the English domain while neglecting the importance of multilingual generalization.
Approach: They propose a multimodal pre-trained model for multilingual document understanding which bridges the language barriers for visually rich document understanding.
Outcome: The proposed model outperforms existing cross-lingual pre-trained models on the XFUND dataset on visual document understanding tasks.
LayoutLMv2: Multi-modal Pre-training for Visually-rich Document Understanding (2021.acl-long)

Copied to clipboard

Challenge: Existing pre-training tasks for text and layout are effective in visually-rich document understanding tasks.
Approach: They propose to combine pre-training tasks with a multi-modal model to model interaction between text, layout and image in a single multi-module framework.
Outcome: The proposed model outperforms LayoutLM by a large margin on visual-rich document understanding tasks.
LayoutReader: Pre-training of Text and Layout for Reading Order Detection (2021.emnlp-main)

Copied to clipboard

Challenge: Existing methods for reading order detection are too laborious to annotate large datasets.
Approach: They propose to use a large-scale dataset to annotate reading order information for document images . they use XML metadata to capture the reading order of WORD documents .
Outcome: The proposed model performs almost perfectly in reading order detection and improves both open-source and commercial OCR engines in ordering text lines in their results.
DocBank: A Benchmark Dataset for Document Layout Analysis (2020.coling-main)

Copied to clipboard

Challenge: Existing approaches for document layout analysis are based on rule-based or machine learning methods that ignore textual information.
Approach: They present a benchmark document layout analysis dataset using a computer vision model . they build strong baselines and manually split train/dev/test sets for evaluation .
Outcome: The proposed model trains on DocBank accurately recognize layout information for a variety of documents.
MANBench: Is Your Multimodal Model Smarter than Human? (2025.findings-acl)

Copied to clipboard

Challenge: Multimodal Large Language Models (MLLMs) have been gaining popularity in multimodal tasks . a bilingual benchmark is available for MLLM users to evaluate their multimodal capabilities .
Approach: They propose a bilingual multimodal ability norms benchmark that measures multimodality across nine tasks.
Outcome: The proposed benchmark compared human performance against state-of-the-art MLLMs.
MMAPG: A Training-Free Framework for Multimodal Multi-hop Question Answering via Adaptive Planning Graphs (2025.emnlp-main)

Copied to clipboard

Challenge: Existing multimodal question answering models rely on sequential retrieval and reasoning, but this single-path paradigm makes them vulnerable to errors due to misleading intermediate steps.
Approach: They propose a multimodal multi-hop question answering framework guided by an Adaptive Planning Graph . they propose modality-specific strategies that dynamically adapt to distinct data types .
Outcome: The proposed framework outperforms existing models that rely on training.
HVGuard: Utilizing Multimodal Large Language Models for Hateful Video Detection (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for hateful video detection rely on unimodal analysis or feature fusion . Existing tools struggle to capture cross-modal interactions and reason through implicit hate in sarcasm and metaphor .
Approach: They propose a reasoning-based hateful video detection framework with multimodal large language models . they integrate Chain-of-Thought reasoning to enhance multimodal interaction modeling .
Outcome: The proposed framework outperforms existing tools on two public datasets covering English and Chinese.
MarkupLM: Pre-training of Text and Markup Language for Visually Rich Document Understanding (2022.acl-long)

Copied to clipboard

Challenge: Existing layout-based pre-training approaches are not easy to apply to VRDU tasks.
Approach: They propose to use markup languages as the backbone for document understanding tasks where text and markup information are jointly pre-trained.
Outcome: The proposed model outperforms existing models on document understanding tasks.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations