mlsanityData to 14 Oct 2026

Does More Retrieved Evidence Help Visual Retrieval-Augmented Generation with Diffusion Language Models?

arXiv 2608.07006 ↗First posted 7 Aug 2026cs.CL cs.CVcited by 0

On the map: Retrieval-augmented generation → RAG robustness and adaptive retrieval

also in 1 more
Abstract (arXiv)

Visual retrieval-augmented generation (RAG) commonly expands the retrieved evidence set to improve answer-page coverage, implicitly assuming that all available evidence should be passed to the generator. We show that this assumption does not hold for diffusion language models (DLMs): retrieving more pages increases answer-page recall, whereas unconditionally passing all retrieved pages to the generator often reduces answer accuracy, primarily because of semantic conflict. A latent-source analysis explains this mismatch through source-coherence loss in parallel denoising, where position-wise proposals can combine incompatible visual sources into unsupported answers. We further find that such interference is already visible in the first-step answer-block distribution, making it possible to assess evidence before decoding. To preserve retrieval coverage while limiting harmful visual exposure, we propose the Entropy-Based Candidate Filter (ECF), a training-free evidence-admission framework. To reduce irrelevant content within individual candidates, ECF constructs multi-granularity evidence units; to identify beneficial additional evidence, it uses blank-controlled block confidence and retrieval rank to determine whether and which candidate should enter the final context. Across three multimodal DLMs and five visual QA benchmarks, ECF improves answer accuracy by 2.62 percentage points on average over the strongest fixed top-k input and, with LLaDA2.0-Uni, by 2.37 percentage points on average over the best competing training-free result for each dataset. These results show that broader retrieval benefits visual DLM-RAG through selective evidence admission rather than unconditional evidence expansion. Code is publicly available at https://github.com/wjkuser/ECF.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in RAG robustness and adaptive retrieval, one of 15 directions in Retrieval-augmented generation. Builds on 17 of its 24 references that are in the map. First posted on arXiv 7 Aug 2026.

Builds on

17 of 24 references.

References · 17

  1. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,455
  2. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  3. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoningcited by 1,479
  4. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,317
  5. 2020Leveraging Passage Retrieval with Generative Models for Open Domain Question Answeringcited by 1,210
  6. 2021Argmax Flows and Multinomial Diffusion: Learning Categorical Distributionscited by 574
  7. 2023Making Retrieval-Augmented Language Models Robust to Irrelevant Contextcited by 369
  8. 2024VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documentscited by 245
  9. 2023Chain-of-Note: Enhancing Robustness in Retrieval-Augmented Language Modelscited by 184
  10. 2025LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuningcited by 148
  11. 2025General-Reasoner: Advancing LLM Reasoning Across All Domainscited by 135
  12. 2025ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLMscited by 43
5 more
  1. 2025Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbonecited by 18
  2. 2025InfoChartQA: A Benchmark for Multimodal Question Answering on Infographic Chartscited by 17
  3. 2025Controllable Logical Hypothesis Generation for Abductive Reasoning in Knowledge Graphscited by 11
  4. 2025Unifying Deductive and Abductive Reasoning in Knowledge Graphs with Masked Diffusion Modelcited by 8
  5. 2026SING: Synthetic Intention Graph for Scalable Active Tool Discovery in LLM Agentscited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Assessing the Downstream Utility of Evidence-Aware Retrieval in RAG
  2. 2026Relevance Is Not Sufficient Evidence: Detecting Evidence Gaps Before Generation in RAG
  3. 2024Investigating Context-Faithfulness in Large Language Models: The Roles of Memory Strength and Evidence Style
  4. 2026Why Retrieval-Augmented Generation Fails: A Graph Perspective
  5. 2026Utility-Oriented Visual Evidence Selection for Multimodal Retrieval-Augmented Generation
  6. 2026Navigating Sparse Evidence: Agentic Visual RAG via Explicit Context Selection and Consolidation
  7. 2026CUE-R: Beyond the Final Answer in Retrieval-Augmented Generation
  8. 2026NeocorRAG: Less Irrelevant Information, More Explicit Evidence, and More Effective Recall via Evidence Chains