DocVQA: A Dataset for VQA on Document Images
By Minesh Mathew, Dimosthenis Karatzas, Raghavan Manmatha, C. V. Jawahar. 2020
1,520 citations recorded by the source corpus.
We present a new dataset for Visual Question Answering (VQA) on document images called DocVQA. The dataset consists of 50,000 questions defined on 12,000+ document images. Detailed analysis of the dataset in comparison with similar datasets for VQA and reading comprehension is presented. We report several baseline results by adopting existing VQA and reading comprehension models. Although the existing models perform reasonably well on certain types of questions, there is large performance gap compared to human performance (94.36% accuracy). The models need to improve specifically on questions where understanding structure of the document is crucial. The dataset, code and leaderboard are available at docvqa.org
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Question Answering & Reading Comprehension
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Mohit Bansal
- Jianfeng Gao
- Luke S. Zettlemoyer
- Quoc V. Le
- Percy Liang
- Yoshua Bengio
- Furu Wei
- Caiming Xiong
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Xiaogang Wang
- Trevor Darrell
- Yu Qiao
- Ming-Hsuan Yang
- Andrew Zisserman
- Jitendra Malik
In Question Answering & Reading Comprehension
Citations in the corpus
- LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention (2023, background): “Alpaca: Five words that rhyme with shock are: rock, pop, shock, cook, and snock.”
- Knowledge Graph Prompting for Multi-Document Question Answering (2023, methodology): “One common solution to overcome this limitation in conventional OD-QA and single document question-answering (D-QA) (Xu et al. 2020; Mathew, Karatzas, and Jawahar 2021) is to retrieve grounding contexts and derive faithful answers from the contexts, i.e., retrieve-and-read (Zhu et al. 2021; Ju et…”
- Qwen-VL: A Frontier Large Vision-Language Model with Versatile Abilities (2023, methodology): “…data for the VQA task which includes GQA (Hudson 2This task is to generate noun/phrase grounded captions (Peng et al., 2023). andManning, 2019), VGQA (Krishna et al., 2017), VQAv2 (Goyal et al., 2017), DVQA (Kafle et al., 2018), OCRVQA (Mishra et al., 2019) and DocVQA (Mathew et al., 2021).”
- TextOCR: Towards large-scale end-to-end reasoning for arbitrary-shaped scene text (2021, background): “A series of datasets were introduced following these which focusing on specific aspects of text-based VQA including OCR-VQA [39], STE-VQA [60], DocVQA [36], PlotQA [37], and LEAF-QA [7].”
- Visual CoT: Advancing Multi-Modal Language Models with a Comprehensive Dataset and Benchmark for Chain-of-Thought Reasoning (2024, methodology): “We choose five text-related datasets to create data in this domain: TextVQA [53], DocVQA [43], DUDE [57], TextCaps [52], SROIE [17].”
- LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model (2023, methodology): “Another example using the OCR expert from DocVQA [44] is presented in Fig.”
- Qwen3-VL Technical Report (2025, positive): “• Document Understanding: We perform comprehensive evaluation on OCR and document understanding ability of Qwen3-VL series across a diverse range OCR related benchmarks: DocVQA (Mathew et al., 2021b), InfoVQA (Mathew et al., 2021a), AI2D (Kembhavi et al., 2016), ChartQA (Masry et al., 2022), OCRBenc”
- ImageBind-LLM: Multi-modality Instruction Tuning (2023, methodology): “We employ 9 benchmarks in the VQA task, namely DocVQA [73], TextVQA [74], STVQA [75], OCR-VQA [76], OKVQA [77], GQA [78], IconQA [79], Visual Spatial Reasoning (VSR) [80], and Visual Dialog (Visdial) [81].”
- Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond (2023, methodology): “We assess our models’ ability toward text-oriented visual question answering on several benchmarks including TextVQA (Sidorov et al., 2020), DocVQA (Mathew et al., 2021), ChartQA (Masry et al., 2022), AI2Diagram (Kembhavi et al., 2016), and OCR-VQA (Mishra et al., 2019).”
- How far are we to GPT-4V? Closing the gap to commercial multimodal models with open-source suites (2024, neutral): “Therefore, we only tested up to 18 tiles maximum, and MMMU was not included when calculating the average score. a considerable margin.”
- CogAgent: A Visual Language Model for GUI Agents (2023, positive): “• Though CogAgent focuses on GUIs, it achieves state-of-the-art generalist performance on nine visual question-answering benchmarks including VQAv2 [1], OK-VQA [23], TextVQA [34], ST-VQA [4], ChartQA [24], infoVQA [26], DocVQA [25], MM-Vet [41], and POPE [19].”
- Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding (2022, background): “DocVQA (Mathew et al., 2021) is a dataset of questions about scanned documents,7 including typewritten, printed, handwritten and born-digital text.”
- LVLM-EHub: A Comprehensive Evaluation Benchmark for Large Vision-Language Models (2023, neutral): “The VQA task includes benchmarks such as DocVQA [41], TextVQA [38], STVQA [79], OCR-VQA [42], OKVQA [80], GQA [43], IconQA [81], Visual Spatial Reasoning (VSR) [82], and WHOOPS [83].”
- MVBench: A Comprehensive Multi-modal Video Understanding Benchmark (2023, methodology): “We include the basic VQA (VQAv2 [23], GQA [27], TGIF-QA [28] and WebVidQA [91]), knowledge-based VQA (OK-VQA [53], AOK-VQA [64] and ViQuAE [36]), OCR-based VQA (OCR-VQA [55], TextVQA [67], ST-VQA [4] and DocVQA [54]), and egocentric VQA from Ego4D [24].”
- InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models (2025, neutral): “Figure 1 provides a detailed assessment of InternVL3’s performance across a diverse set of benchmarks, including MMMU [141], MathVista [80], AI2D [57], ChartQA [91], DocVQA [93], InfographicVQA [92], HallusionBench [45], OCRBench [76], and LongVideoBench [129].”
- StructuralLM: Structural Pre-training for Form Understanding (2021, background): “DocVQA (Mathew et al., 2020) is a VQA dataset on the scanned document understanding field.”
- WebSRC: A Dataset for Web-Based Structural Reading Comprehension (2021, background): “Text VQA (Mishra et al., 2019; Singh et al., 2019; Mathew et al., 2021) is a kind of VQA (visual question answering) task (Antol et al., 2015), whose task is to answer questions about a real-world image, and questions in this task are about the texts in the image.”
- Monkey: Image Resolution and Text Label are Important Things for Large Multi-Modal Models (2023, neutral): “…18k GQA [20] 150k ScienceQA [33] 18k VizWiz [17] 20k Scene Text-centric VQA TextVQA [45] 34k OCRVQA [38] 250k AI2D [23] 24k Doc-oriented VQA DocVQA [36] 118k ChartQA [35] 84k InfoVQA [37] 47k DeepForm [47] 7k KLC [46] 27k WTQ [40] 28k TabFact [8] 91k VisualMRC [48] 21k individually described by…”