Collapse, Not Complexity: Failure-Conditioned Decomposition Repair for End-to-End Document Parsing
Not on the map. Why?
Abstract (arXiv)
End-to-end document parsers increasingly offer an optional reasoning mode for complex pages. On a 180-page entropy-stratified discovery sample with one frozen 4B checkpoint, complexity is the wrong decision variable. Reasoning lowers mean quality by 2.21 Overall at 1.54x tokens; a preregistered input-only model cannot predict its signed benefit (held-out AUROC 0.47, indistinguishable from chance). The benefit concentrates on pages whose ordinary pass has already collapsed, and they do not look complex: shared collapses have lower layout entropy than healthy ones yet consume 19x the tokens as degenerate repetition that doubling the budget does not cure. Switching modes rarely repairs them: 83% recur under reasoning. We instead detect collapse from the ordinary-pass trace, decompose the page by projection, and re-parse each region. Repair gains 1.40 Overall (95% CI [0.68, 2.16]) at 1.13x tokens, replicates across three checkpoints, and, with all parameters frozen, gains 2.41 (CI [1.64, 3.46]) on the remaining 1,175 benchmark pages.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 22 of its 27 references that are in the map. First posted on arXiv 20 Sep 2026.
Builds on
22 of 27 references.
References · 22
- 2019Image-based table recognition: data, model, and evaluationcited by 233
- 2025MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipecited by 174
- 2024OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotationscited by 144
- 2025AdaptThink: Reasoning Models Can Learn When to Thinkcited by 141
- 2025MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsingcited by 108
- 2024Mind Your Step (by Step): Chain-of-Thought can Reduce Performance on Tasks where Thinking Makes Humans Worsecited by 107
- 2025PaddleOCR-VL: Boosting Multilingual Document Parsing via a 0.9B Ultra-Compact Vision-Language Modelcited by 84
- 2025Prolonged Reasoning Is Not All You Need: Certainty-Based Adaptive Routing for Efficient LLM/MLLM Reasoningcited by 34
- 2025Think or Not? Selective Reasoning via Reinforcement Learning for Vision-Language Modelscited by 30
- 2026GLM-OCR Technical Reportcited by 24
- 2026Qianfan-OCR: A Unified End-to-End Model for Document Intelligencecited by 20
- 2025Consensus Entropy: Harnessing Multi-VLM Agreement for Self-Verifying and Self-Improving OCRcited by 12
10 more
- 2025Mixture-of-Visual-Thoughts: Exploring Context-Adaptive Reasoning Mode Selection for General Visual Reasoningcited by 7
- 2025Reading or Reasoning? Format Decoupled Reinforcement Learning for Document OCRcited by 7
- 2025DOCR-Inspector: Fine-Grained and Automated Evaluation of Document Parsing with VLMcited by 6
- 2026Addressing Overthinking in Large Vision-Language Models via Gated Perception-Reasoning Optimizationcited by 6
- 2025The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Modelscited by 4
- 2026How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settingscited by 4
- 2026Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Docited by 3
- 2026Multi-Expert Routing for Multi-Domain Low-Resource OCR: A Manchu Case Studycited by 1
- 2026LayoutLite: Token-Level Implicit Layout Analysis for Efficient Document OCRcited by 1
- 2026Pre-Inference Routing for Cost-Efficient Document Field Extractioncited by 1
Built on by
Not cited yet.