Research map / Vision-language and multimodal / Visual question answering
VinQA: Visual Elements Interleaved Long-form Answer Generation for Real-World Multimodal Document QA
Young Rok Jang, Hyesoo Kong, Kyunghwan An, Jae Sub Huh, Gyeonghun KIM, Stanley Jungkyu Choi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune on VinQA training split | Qwen2.5-VL-7B (Page Encoding) · VinQA test | M-GroSE Avg: 1.99 -> 3.34 | Improved |
| fine-tune on VinQA training split | Qwen2.5-VL-7B (Page Encoding) · VinQA test | Visual Source F1: 0.31 -> 0.55 | Improved |
| fine-tune on VinQA training split | Qwen2.5-VL-7B (Modality Encoding) · VinQA test | M-GroSE Avg: 2.08 -> 3.33 | Improved |
| fine-tune on VinQA training split | Qwen2.5-VL-7B (Modality Encoding) · VinQA test | Visual Source F1: 0.40 -> 0.58 | Improved |
| use Modality Encoding instead of Page Encoding | GPT-4.1 · VinQA test | M-GroSE Avg: 3.46 -> 3.63 | Improved |
| use Modality Encoding instead of Page Encoding | Gemini 2.0 Flash · VinQA test | M-GroSE Avg: 2.86 -> 3.43 | Improved |
| use Modality Encoding instead of Page Encoding | Claude 3.5 Sonnet · VinQA test | Visual Source F1: 0.65 -> 0.70 | Improved |
| use Modality Encoding instead of Page Encoding | Qwen2.5-VL-7B fine-tuned on VinQA · VinQA test | M-GroSE Avg: 3.34 -> 3.33 | No clear effect |
| use Modality Encoding instead of Page Encoding | Qwen2.5-VL-7B fine-tuned on VinQA · VinQA test | Visual Source F1: 0.55 -> 0.58 | No clear effect |
| fine-tune on VinQA training split | Qwen2.5-VL-7B (Page Encoding) · Visual G-Eval Effectiveness | Visual G-Eval: 3.44 -> 3.94 | Improved |
| fine-tune on VinQA training split | Qwen2.5-VL-7B (Modality Encoding) · Visual G-Eval Effectiveness | Visual G-Eval: 4.00 -> 4.17 | Improved |
| fine-tune on VinQA training split | Qwen2.5-VL-7B (Page Encoding) · Visual G-Eval Position | Visual G-Eval: 4.24 -> 4.77 | Improved |
Closest papers by meaning
- Benchmarking Retrieval-Augmented Multimomal Generation for Document Question Answering (NeurIPS 2025 · similarity 0.91)
- SPIQA: A Dataset for Multimodal Question Answering on Scientific Papers (NeurIPS 2024 · similarity 0.88)
- GRAM: Global Reasoning for Multi-Page VQA (CVPR 2024 · similarity 0.88)
- Marten: Visual Question Answering with Mask Generation for Multi-modal Document Understanding (CVPR 2025 · similarity 0.88)
- InfoChartQA: A Benchmark for Multimodal Question Answering on Infographic Charts (NeurIPS 2025 · similarity 0.88)
- UPME: An Unsupervised Peer Review Framework for Multimodal Large Language Model Evaluation (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
