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

arXiv: https://arxiv.org/abs/2608.28698 · HTML: https://mlsanity.com/paper/2608.28698/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Document understanding and multimodal RAG, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 24 of its 32 references that are in the map.
- First posted on arXiv 27 Aug 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Document understanding and multimodal RAG](https://mlsanity.com/mllm/document-understanding-and-multimodal-rag/)

## Builds on (most cited first, 24 of 24)

- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://mlsanity.com/paper/2010.04159/) (2020)
- [Longformer: The Long-Document Transformer](https://mlsanity.com/paper/2004.05150/) (2020)
- [Towards VQA Models That Can Read](https://mlsanity.com/paper/1904.08920/) (2019)
- [Big Bird: Transformers for Longer Sequences](https://mlsanity.com/paper/2007.14062/) (2020)
- [Reformer: The Efficient Transformer](https://mlsanity.com/paper/2001.04451/) (2020)
- [DocVQA: A Dataset for VQA on Document Images](https://mlsanity.com/paper/2007.00398/) (2020)
- [LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention](https://mlsanity.com/paper/2303.16199/) (2023)
- [Efficient Content-Based Sparse Attention with Routing Transformers](https://mlsanity.com/paper/2003.05997/) (2020)
- [MinerU: An Open-Source Solution for Precise Document Content Extraction](https://mlsanity.com/paper/2409.18839/) (2024)
- [Nougat: Neural Optical Understanding for Academic Documents](https://mlsanity.com/paper/2308.13418/) (2023)
- [Image-based table recognition: data, model, and evaluation](https://mlsanity.com/paper/1911.10683/) (2019)
- [Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution](https://mlsanity.com/paper/2307.06304/) (2023)
- [General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model](https://mlsanity.com/paper/2409.01704/) (2024)
- [OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations](https://mlsanity.com/paper/2412.07626/) (2024)
- [DocLayNet: A Large Human-Annotated Dataset for Document-Layout Analysis](https://mlsanity.com/paper/2206.01062/) (2022)
- [Global Table Extractor (GTE): A Framework for Joint Table Identification and Cell Structure Recognition Using Visual Context](https://mlsanity.com/paper/2005.00589/) (2020)
- [MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm](https://mlsanity.com/paper/2506.05218/) (2025)
- [DocLayout-YOLO: Enhancing Document Layout Analysis through Diverse Synthetic Data and Global-to-Local Adaptive Perception](https://mlsanity.com/paper/2410.12628/) (2024)
- [Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting](https://mlsanity.com/paper/2505.14059/) (2025)
- [DeepSeek-OCR 2: Visual Causal Flow](https://mlsanity.com/paper/2601.20552/) (2026)
- [Vision Grid Transformer for Document Layout Analysis](https://mlsanity.com/paper/2308.14978/) (2023)
- [DocVLM: Make Your VLM an Efficient Reader](https://mlsanity.com/paper/2412.08746/) (2024)
- [Hierarchical Visual Feature Aggregation for OCR-Free Document Understanding](https://mlsanity.com/paper/2411.05254/) (2024)

## Abstract (arXiv)

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.