M-DocSum: Do LVLMs Genuinely Comprehend Interleaved Image-Text in Document Summarization?

• URL: http://arxiv.org/abs/2503.21839v1
• Date: Thu, 27 Mar 2025 07:28:32 GMT
• Title: M-DocSum: Do LVLMs Genuinely Comprehend Interleaved Image-Text in Document Summarization?
• Authors: Haolong Yan, Kaijun Tan, Yeqing Shen, Xin Huang, Zheng Ge, Xiangyu Zhang, Si Li, Daxin Jiang,
• Abstract summary: We investigate whether Large Vision-Language Models (LVLMs) genuinely comprehend interleaved image-text in the document.<n>Existing document understanding benchmarks often assess LVLMs using question-answer formats.<n>We introduce a novel and challenging Multimodal Document Summarization Benchmark (M-DocSum-Bench)<n>M-DocSum-Bench comprises 500 high-quality arXiv papers, along with interleaved multimodal summaries aligned with human preferences.
• Score: 49.53982792497275
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We investigate a critical yet under-explored question in Large Vision-Language Models (LVLMs): Do LVLMs genuinely comprehend interleaved image-text in the document? Existing document understanding benchmarks often assess LVLMs using question-answer formats, which are information-sparse and difficult to guarantee the coverage of long-range dependencies. To address this issue, we introduce a novel and challenging Multimodal Document Summarization Benchmark (M-DocSum-Bench), which comprises 500 high-quality arXiv papers, along with interleaved multimodal summaries aligned with human preferences. M-DocSum-Bench is a reference-based generation task and necessitates the generation of interleaved image-text summaries using provided reference images, thereby simultaneously evaluating capabilities in understanding, reasoning, localization, and summarization within complex multimodal document scenarios. To facilitate this benchmark, we develop an automated framework to construct summaries and propose a fine-grained evaluation method called M-DocEval. Moreover, we further develop a robust summarization baseline, i.e., M-DocSum-7B, by progressive two-stage training with diverse instruction and preference data. The extensive results on our M-DocSum-Bench reveal that the leading LVLMs struggle to maintain coherence and accurately integrate information within long and interleaved contexts, often exhibiting confusion between similar images and a lack of robustness. Notably, M-DocSum-7B achieves state-of-the-art performance compared to larger and closed-source models (including GPT-4o, Gemini Pro, Claude-3.5-Sonnet and Qwen2.5-VL-72B, etc.), demonstrating the potential of LVLMs for improved interleaved image-text understanding. The code, data, and models are available at https://github.com/stepfun-ai/M-DocSum-Bench.

Related papers

• MCiteBench: A Benchmark for Multimodal Citation Text Generation in MLLMs [31.793037002996257]
  Multimodal Large Language Models (MLLMs) have advanced in integrating diverse modalities but frequently suffer from hallucination.<n>Existing work primarily focuses on generating citations for text-only content, overlooking the challenges and opportunities of multimodal contexts.<n>We introduce MCiteBench, the first benchmark designed to evaluate and analyze the multimodal citation text generation ability of MLLMs.
  arXiv  Detail & Related papers  (2025-03-04T13:12:39Z)
• Towards Text-Image Interleaved Retrieval [49.96332254241075]
  We introduce the text-image interleaved retrieval (TIIR) task, where the query and document are interleaved text-image sequences.<n>We construct a TIIR benchmark based on naturally interleaved wikiHow tutorials, where a specific pipeline is designed to generate interleaved queries.<n>We propose a novel Matryoshka Multimodal Embedder (MME), which compresses the number of visual tokens at different granularity.
  arXiv  Detail & Related papers  (2025-02-18T12:00:47Z)
• HERA: Improving Long Document Summarization using Large Language Models with Context Packaging and Reordering [6.876612430571396]
  We propose a novel summary generation framework, called HERA.<n>We first segment a long document by its semantic structure and retrieve text segments about the same event, and finally reorder them to form the input context.<n>The experimental results show that HERA outperforms foundation models in ROUGE, BERTScore and faithfulness metrics.
  arXiv  Detail & Related papers  (2025-02-01T14:55:06Z)
• VisDoM: Multi-Document QA with Visually Rich Elements Using Multimodal Retrieval-Augmented Generation [100.06122876025063]
  This paper introduces VisDoMBench, the first comprehensive benchmark designed to evaluate QA systems in multi-document settings.<n>We propose VisDoMRAG, a novel multimodal Retrieval Augmented Generation (RAG) approach that simultaneously utilizes visual and textual RAG.
  arXiv  Detail & Related papers  (2024-12-14T06:24:55Z)
• M3DocRAG: Multi-modal Retrieval is What You Need for Multi-page Multi-document Understanding [63.33447665725129]
  We introduce M3DocRAG, a novel multi-modal RAG framework that flexibly accommodates various document contexts. M3DocRAG can efficiently handle single or many documents while preserving visual information. We also present M3DocVQA, a new benchmark for evaluating open-domain DocVQA over 3,000+ PDF documents with 40,000+ pages.
  arXiv  Detail & Related papers  (2024-11-07T18:29:38Z)
• DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding [40.38251904765156]
  Text-rich document understanding (TDU) requires comprehensive analysis of documents containing substantial textual content and complex layouts.<n>We introduce DocLayLLM, an efficient multi-modal extension of Multimodal Large Language Models (MLLMs) specifically designed for TDU.
  arXiv  Detail & Related papers  (2024-08-27T13:13:38Z)
• DOCBENCH: A Benchmark for Evaluating LLM-based Document Reading Systems [99.17123445211115]
  We introduce DocBench, a benchmark to evaluate large language model (LLM)-based document reading systems. Our benchmark involves the recruitment of human annotators and the generation of synthetic questions. It includes 229 real documents and 1,102 questions, spanning across five different domains and four major types of questions.
  arXiv  Detail & Related papers  (2024-07-15T13:17:42Z)
• TextHawk: Exploring Efficient Fine-Grained Perception of Multimodal Large Language Models [9.232693392690702]
  TextHawk is a document-oriented Multimodal Large Language Model (MLLM) It is designed to explore efficient fine-grained perception by designing four dedicated components. We conduct extensive experiments on both general and document-oriented MLLM benchmarks, and show that TextHawk outperforms the state-of-the-art methods.
  arXiv  Detail & Related papers  (2024-04-14T09:48:37Z)
• SEED-Bench-2: Benchmarking Multimodal Large Language Models [67.28089415198338]
  Multimodal large language models (MLLMs) have recently demonstrated exceptional capabilities in generating not only texts but also images given interleaved multimodal inputs. SEED-Bench-2 comprises 24K multiple-choice questions with accurate human annotations, which spans 27 dimensions. We evaluate the performance of 23 prominent open-source MLLMs and summarize valuable observations.
  arXiv  Detail & Related papers  (2023-11-28T05:53:55Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.