mlsanityData to 14 Oct 2026

Capability-Routed Visual Retrieval and Evidence Threading for Long-Context Document Question Answering

arXiv 2609.13268 ↗First posted 7 Sep 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Annual reports, diligence packs, and infographic dashboards bury numbers in page images: axes, cell grids, and footnotes that OCR pipelines flatten and that page-level visual retrievers still treat as interchangeable in-context examples. We keep a frozen Qwen2.5-VL-7B-Instruct generator and a ColPali / VisRAG-Ret page index, and insert three modules. A capability-aware visual router (CAVR) tags each retrieved page as text, table, chart, layout, or mixed and mixes specialist experts before generation. Weak-to-strong page selection (WSPS) distils a frozen 7B answerability teacher into a 3B selection head so ranking is no longer a single InfoNCE score. Visual evidence threading (VET) builds layout-anchored paths of length at most three and lets the generator read the thread rather than a flat top-k list. On gold-page DocVQA / ChartQA / InfographicVQA the 7B system reaches 96.3 / 90.1 / 85.4. Under the VisRAG top-3 protocol the mean generation accuracy is 62.74 versus 59.39 for the same backbone with concatenation. On MMLongBench-Doc retrieve-then-read, F1 moves from 19.2 to 22.6 and multi-page accuracy from 16.4 to 21.2. ViDoRe nDCG@5 after WSPS reranking is 83.6, with TAT-DQA financial reports at 70.4.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 36 of its 49 references that are in the map. First posted on arXiv 7 Sep 2026.

Builds on

36 of 49 references.

References · 36

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  3. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
  4. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,824
  5. 2023Sigmoid Loss for Language Image Pre-Trainingcited by 3,109
  6. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,565
  7. 2024Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scalingcited by 1,599
  8. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoningcited by 1,479
  9. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,317
  10. 2020ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERTcited by 1,178
  11. 2023CogAgent: A Visual Language Model for GUI Agentscited by 704
  12. 2021InfographicVQAcited by 437
24 more
  1. 2021OCR-free Document Understanding Transformercited by 407
  2. 2022Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understandingcited by 366
  3. 2023ARES: An Automated Evaluation Framework for Retrieval-Augmented Generation Systemscited by 265
  4. 2024VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documentscited by 245
  5. 2024mPLUG-DocOwl 1.5: Unified Structure Learning for OCR-free Document Understandingcited by 220
  6. 2023Can Pre-trained Vision and Language Models Answer Visual Information-Seeking Questions?cited by 207
  7. 2024ColPali: Efficient Document Retrieval with Vision Language Modelscited by 195
  8. 2022MatCha: Enhancing Visual Language Pretraining with Math Reasoning and Chart Derenderingcited by 133
  9. 2024Visual In-Context Learning for Large Vision-Language Modelscited by 128
  10. 2021KAT: A Knowledge Augmented Transformer for Vision-and-Languagecited by 126
  11. 2022Retrieval Augmented Visual Question Answering with Outside Knowledgecited by 94
  12. 2023Thread of Thought Unraveling Chaotic Contextscited by 72
  13. 2023Modular Visual Question Answering via Code Generationcited by 48
  14. 2024Rationale-Guided Retrieval Augmented Generation for Medical Question Answeringcited by 41
  15. 2023Retrieval-augmented Image Captioningcited by 29
  16. 2024GATEAU: Selecting Influential Samples for Long Context Alignmentcited by 20
  17. 2025A Goal Without a Plan Is Just a Wish: Efficient and Effective Global Planner Training for Long-Horizon Agent Taskscited by 19
  18. 2025Teaching Large Language Models to Maintain Contextual Faithfulness via Synthetic Tasks and Reinforcement Learningcited by 11
  19. 2025LAD-RAG: Layout-aware Dynamic RAG for Visually-Rich Document Understandingcited by 7
  20. 2022Double Retrieval and Ranking for Accurate Question Answeringcited by 4
  21. 2026MARA: A Multimodal Adaptive Retrieval-Augmented Framework for Document Question Answeringcited by 3
  22. 2026MAGE-RAG: Multigranular Adaptive Graph Evidence for Agentic Multimodal RAG in Long-Document QAcited by 2
  23. 2026AnchorFold: A Focus-Then-Fold Framework via Recursive Attention Propagation for Efficient Multi-Vector Visual Document Retrievalcited by 1
  24. 2026PULSAR: Pooled Unified Late-Interaction Search and Retrieval for Enterprise Visual Document RAGcited by 1

Built on by

Not cited yet.