State-Conditioned Visual Evidence Retrieval for Fine-Grained Perception in Document Vision-Language Models

• URL: http://arxiv.org/abs/2608.28698v1
• Date: Thu, 27 Aug 2026 12:56:45 GMT
• Title: State-Conditioned Visual Evidence Retrieval for Fine-Grained Perception in Document Vision-Language Models
• Abstract summary: Document parsing places stronger demands on fine-grained visual perception.<n>Existing vision-language model (VLM)-based parsing approaches rely on globally compressed visual tokens.<n>We formulate perception as state-conditioned visual evidence retrieval (SCVER) during autoregressive decoding.
• Score: 49.57973975600996
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Compared with typical vision-language tasks, document parsing places stronger demands on fine-grained visual perception. Existing vision-language model (VLM)-based parsing approaches rely on globally compressed visual tokens, where fine-grained details are entangled within a single representation and repeatedly accessed during decoding. However, we observe that the visual evidence for each prediction is typically localized and conditioned on the current decoding state, whereas such representations must be accessed in full at every decoding step, resulting in inefficient computation. To address this mismatch, we formulate perception as state-conditioned visual evidence retrieval (SCVER) during autoregressive decoding. The model operates on a compact global representation for coarse structure and retrieves a small set of relevant high-resolution regions conditioned on the current token state. This coarse-to-fine design enables on-demand access to fine-grained visual cues, relieving globally shared representations from encoding all fine-grained details. We further find that learning such state-conditioned retrieval in VLMs is challenging and unstable. To stabilize this process, we introduce a Spatially-Guided Learning Objective (SGLO) to guide the retrieval process. Experiments on document parsing benchmarks show that SCVER improves robustness under reduced input resolution and achieves a better accuracy-efficiency trade-off, demonstrating the effectiveness of on-demand visual evidence retrieval for fine-grained perception.

Related papers

• FOVEA: Focused On-Demand Visual Evidence Adaptation for Cache-Friendly Multimodal Speculative Decoding [42.559533294854134]
  Multimodal speculative decoding accelerates vision-language models by allowing a lightweight draft model to propose candidate tokens for parallel verification by a larger target model.<n>Existing methods typically condition the drafter on a fixed visual interface, such as a predefined visual-token budget or a static compressed representation.<n>We propose FOVEA, a cache-friendly approach that builds a reusable visual memory and dynamically retrieves a bounded subset for a draft state.
  arXiv  Detail & Related papers  (2026-08-24T07:09:02Z)
• ViQ: Text-Aligned Visual Quantized Representations at Any Resolution [91.46185855575789]
  We present ViQ, a Visual Quantized Representations framework, which is designed to balance semantics and details in discrete representations.<n>Our approach structures quantization learning into two stages: text-aligned pre-training and feature discretization.<n>Experiments on multimodal tasks demonstrate that ViQ achieves competitive performance compared to state-of-the-art multimodal vision encoders.
  arXiv  Detail & Related papers  (2026-06-25T17:29:27Z)
• MURE: Hierarchical Multi-Resolution Encoding via Vision-Language Models for Visual Document Retrieval [111.99106496142474]
  Visual Document Retrieval (VDR) requires representations that capture both fine-grained visual details and global document structure.<n>Existing VDR models struggle to balance effectiveness and efficiency when processing high-resolution documents.<n>We propose a new X-VisEmb paradigm that progresses from multi-resolution sampling and encoding, through cross-granularity feature fusion, to adaptive representation distillation.
  arXiv  Detail & Related papers  (2026-03-07T15:17:22Z)
• Visual Self-Refine: A Pixel-Guided Paradigm for Accurate Chart Parsing [76.2602505940467]
  Existing models often struggle with visually dense charts, leading to errors like data omission, misalignment, and hallucination.<n>Inspired by the human strategy of using a finger as a visual anchor'' to ensure accuracy when reading complex charts, we propose a new paradigm named Visual Self-Refine (VSR)<n>The core idea of VSR is to enable a model to generate pixel-level localization outputs, visualize them, and then feed these visualizations back to itself, allowing it to intuitively inspect and correct its own potential visual perception errors.
  arXiv  Detail & Related papers  (2026-02-18T13:40:53Z)
• VISTA-Bench: Do Vision-Language Models Really Understand Visualized Text as Well as Pure Text? [51.02924254085878]
  Vision-Language Models (VLMs) have achieved impressive performance in cross-modal understanding across textual and visual inputs.<n>We introduce VISTA-Bench, a benchmark from multimodal perception, reasoning, to unimodal understanding domains.
  arXiv  Detail & Related papers  (2026-02-04T17:48:55Z)
• Revisit What You See: Disclose Language Prior in Vision Tokens for LVLM Decoding [6.612630497074871]
  Large Vision-Language Models (LVLMs) achieve strong performance across multimodal tasks by integrating visual perception with language understanding.<n>We propose ReVisiT, a training-free decoding method that references vision tokens to guide text generation.
  arXiv  Detail & Related papers  (2025-06-11T08:46:55Z)
• GlocalCLIP: Object-agnostic Global-Local Prompt Learning for Zero-shot Anomaly Detection [5.530212768657544]
  We introduce glocal contrastive learning to improve the complementary learning of global and local prompts.<n>The generalization performance of GlocalCLIP in ZSAD was demonstrated on 15 real-world datasets.
  arXiv  Detail & Related papers  (2024-11-09T05:22:13Z)
• Calibrated Self-Rewarding Vision Language Models [27.686545023186852]
  Large Vision-Language Models (LVLMs) have made substantial progress by integrating pre-trained large language models (LLMs) and vision models through instruction tuning. LVLMs often exhibit the hallucination phenomenon, where generated text responses appear linguistically plausible but contradict the input image. We propose the Calibrated Self-Rewarding (CSR) approach, which enables the model to self-improve by iteratively generating candidate responses, evaluating the reward for each response, and curating preference data for fine-tuning.
  arXiv  Detail & Related papers  (2024-05-23T14:30:33Z)
• Zero-Shot Video Moment Retrieval from Frozen Vision-Language Models [58.17315970207874]
  We propose a zero-shot method for adapting generalisable visual-textual priors from arbitrary VLM to facilitate moment-text alignment. Experiments conducted on three VMR benchmark datasets demonstrate the notable performance advantages of our zero-shot algorithm.
  arXiv  Detail & Related papers  (2023-09-01T13:06:50Z)

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.