AI Paper Atlas

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

CVPR 2026 · Vision-language and multimodal › Visual question answering · approach: Multimodal question answering datasets · 1 citations (OpenAlex)

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 changedWhereMeasuredEffect
fine-tune on VinQA training splitQwen2.5-VL-7B (Page Encoding) · VinQA testM-GroSE Avg: 1.99 -> 3.34Improved
fine-tune on VinQA training splitQwen2.5-VL-7B (Page Encoding) · VinQA testVisual Source F1: 0.31 -> 0.55Improved
fine-tune on VinQA training splitQwen2.5-VL-7B (Modality Encoding) · VinQA testM-GroSE Avg: 2.08 -> 3.33Improved
fine-tune on VinQA training splitQwen2.5-VL-7B (Modality Encoding) · VinQA testVisual Source F1: 0.40 -> 0.58Improved
use Modality Encoding instead of Page EncodingGPT-4.1 · VinQA testM-GroSE Avg: 3.46 -> 3.63Improved
use Modality Encoding instead of Page EncodingGemini 2.0 Flash · VinQA testM-GroSE Avg: 2.86 -> 3.43Improved
use Modality Encoding instead of Page EncodingClaude 3.5 Sonnet · VinQA testVisual Source F1: 0.65 -> 0.70Improved
use Modality Encoding instead of Page EncodingQwen2.5-VL-7B fine-tuned on VinQA · VinQA testM-GroSE Avg: 3.34 -> 3.33No clear effect
use Modality Encoding instead of Page EncodingQwen2.5-VL-7B fine-tuned on VinQA · VinQA testVisual Source F1: 0.55 -> 0.58No clear effect
fine-tune on VinQA training splitQwen2.5-VL-7B (Page Encoding) · Visual G-Eval EffectivenessVisual G-Eval: 3.44 -> 3.94Improved
fine-tune on VinQA training splitQwen2.5-VL-7B (Modality Encoding) · Visual G-Eval EffectivenessVisual G-Eval: 4.00 -> 4.17Improved
fine-tune on VinQA training splitQwen2.5-VL-7B (Page Encoding) · Visual G-Eval PositionVisual G-Eval: 4.24 -> 4.77Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy