mlsanityData to 7 Oct 2026

ShredBench: Evaluating the Semantic Reasoning Capabilities of Multimodal LLMs in Document Reconstruction

arXiv 2604.23813 ↗First posted 26 Apr 2026cs.CV cs.CLcited by 0

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Multimodal Large Language Models (MLLMs) have achieved remarkable performance in Visually Rich Document Understanding (VRDU) tasks, but their capabilities are mainly evaluated on pristine, well-structured document images. We consider content restoration from shredded fragments, a challenging VRDU setting that requires integrating visual pattern recognition with semantic reasoning under significant content discontinuities. To facilitate systematic evaluation of complex VRDU tasks, we introduce ShredBench, a benchmark supported by an automated generation pipeline that renders fragmented documents directly from Markdown. The proposed pipeline ensures evaluation validity by allowing the flexible integration of latest or unseen textual sources to prevent training data contamination. ShredBench assesses four scenarios (English, Chinese, Code, Table) with three fragmentation granularities (8, 12, 16 pieces). Empirical evaluations on state-of-the-art MLLMs reveal a significant performance gap: The method is effective on intact documents; however, once the document is shredded, restoration becomes a significant challenge, with NED dropping sharply as fragmentation increases. Our findings highlight that current MLLMs lack the fine-grained cross-modal reasoning required to bridge visual discontinuities, identifying a critical gap in robust VRDU research.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 31 of its 66 references that are in the map. First posted on arXiv 26 Apr 2026.

Builds on

31 of 66 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024Read and Think: An Efficient Step-wise Multimodal Language Model for Document Understanding and Reasoning26 Feb
  2. 2023MMBench: Is Your Multi-modal Model an All-around Player?12 Jul

Other references · 29

  1. 2023Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyondcited by 2,179
  2. 2024ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Toolscited by 1,096
  3. 2023SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehensioncited by 929
  4. 2023A Survey on Multimodal Large Language Modelscited by 898
  5. 2023CogVLM: Visual Expert for Pretrained Language Modelscited by 704
  6. 2024DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understandingcited by 559
  7. 2023Nougat: Neural Optical Understanding for Academic Documentscited by 240
  8. 2023KOSMOS-2.5: A Multimodal Literate Modelcited by 95
  9. 2023Vary: Scaling up the Vision Vocabulary for Large Vision-Language Modelscited by 84
  10. 2025VRAG-RL: Empower Vision-Perception-Based RAG for Visually Rich Information Understanding via Iterative Reasoning with Reinforcement Learningcited by 68
  11. 2025ViDoRAG: Visual Document Retrieval-Augmented Generation via Dynamic Iterative Reasoning Agentscited by 64
  12. 2025Magistralcited by 57
17 more
  1. 2024From Code to Correctness: Closing the Last Mile of Code Generation with Hierarchical Debuggingcited by 46
  2. 2025LongCodeZip: Compress Long Context for Code Language Modelscited by 38
  3. 2025SWE-QA: Can Language Models Answer Repository-level Code Questions?cited by 37
  4. 2024M-Longdoc: A Benchmark For Multimodal Super-Long Document Understanding And A Retrieval-Aware Tuning Frameworkcited by 36
  5. 2025Pruning the Unsurprising: Efficient LLM Reasoning via First-Token Surprisalcited by 32
  6. 2025WildDoc: How Far Are We from Achieving Comprehensive and Robust Document Understanding in the Wild?cited by 26
  7. 2025HunyuanOCR Technical Reportcited by 18
  8. 2026In Line with Context: Repository-Level Code Generation via Context Inliningcited by 18
  9. 2026Seeing is Coding: On the Effectiveness of Vision Language Models in Code Understandingcited by 13
  10. 2025Jigsaw-Puzzles: From Seeing to Understanding to Reasoning in Vision-Language Modelscited by 9
  11. 2026EffiSkill: Agent Skill Based Automated Code Efficiency Optimizationcited by 9
  12. 2025DocPTBench: Benchmarking End-to-End Photographed Document Parsing and Translationcited by 8
  13. 2026Tears or Cheers? Benchmarking LLMs via Culturally Elicited Distinct Affective Responsescited by 6
  14. 2025Progressive Supernet Training for Efficient Visual Autoregressive Modelingcited by 5
  15. 2023PairingNet: A Learning-based Pair-searching and -matching Network for Image Fragmentscited by 4
  16. 2026Readability-Robust Code Summarization via Meta Curriculum Learningcited by 4
  17. 2026Who Wrote This Line? Evaluating the Detection of LLM-Generated Classical Chinese Poetrycited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025DocSLM: A Small Vision-Language Model for Long Multimodal Document Understanding
  2. 2026DeCoRAG: Cognitive Decoupling and Semantic-Aware Cropping for Complex Document Understanding
  3. 2025Document Haystack: A Long Context Multimodal Image/Document Understanding Vision LLM Benchmark
  4. 2024MMDocBench: Benchmarking Large Vision-Language Models for Fine-Grained Visual Document Understanding
  5. 2024SV-RAG: LoRA-Contextualizing Adaptation of MLLMs for Long Document Understanding
  6. 2026Unlocking Multimodal Document Intelligence: From Current Triumphs to Future Frontiers of Visual Document Retrieval
  7. 2026Real5-OmniDocBench: A Full-Scale Physical Reconstruction Benchmark for Robust Document Parsing in the Wild
  8. 2026SynthDocBench: Controlled Benchmark for Long-Context Visual Document Understanding