FinixDoc: Rethinking Financial Document Parsing Beyond Saturated Benchmarks
2026-08-24 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors developed FinixDoc, a system designed to accurately read and understand financial documents, which can be tricky because these documents vary in quality and size. They created a special vision-language model called FinixDoc-VL and trained it using techniques that help it learn financial text even when the quality is low. To check performance, they introduced a new way to evaluate systems based on how documents look and how big they are, and they made a set of test documents called FinixDocBench. Their system performed better than other open-source models, especially on complex internal financial documents.
financial document parsingvision-language modelcontrastive learningreinforcement learningdocument qualityFinixDocBenchagentic parsinghuman-in-the-loopdata pipelinecompliance review
Authors
Hang Wang, Jin Zhang, Guoliang Xu, Pengyue Lu, Yao Li, Zijiao Zhang, Tianyu Huang, Weiqi Xiong, Yulong Wang, Chuqiao Lu, Wenkang Huang, Kai Yang, Yadong Li, Hui Li, Xingzhong Xu, Xiao Xu
Abstract
Financial document parsing requires accuracy, structural consistency, and verifiability that current benchmarks often fail to reflect. We present FinixDoc, an end-to-end agentic parsing system for real-world financial documents, with FinixDoc-VL, a 4B-scale vision-language model built on Qwen3-VL-4B, as its core parser. To characterize the gap between benchmark and deployment performance, we introduce a Document Parsing Capability Matrix organized along two practical axes: visual quality and document scale. Guided by this matrix, FinixDoc-VL is trained with a domain-adapted recipe combining homoglyph-aware contrastive learning and multi-stage reinforcement learning with composite domain-specific rewards. To better leverage our accumulated advantage in low-quality financial-document data and support large-scale, high-quality data production, we further build a human-in-the-loop Data Factory pipeline with confidence-aware expert review. For evaluation, we construct FinixDocBench, a financial-domain evaluation suite covering digital-native, camera-captured, ultra-large-page, and internal-workflow scenarios, with a compliance-reviewed subset released alongside this technical report. On its main subsets, FinixDoc-VL achieves the highest overall score (81.43) among evaluated baselines, outperforming the next-best open-source model by 5.13 points, with the largest gains on internal financial workflows (FinixInner: 84.08 vs. 78.73).