Research map / Vision-language and multimodal / Visual question answering
Ask Questions With Double Hints: Visual Question Generation With Answer-Awareness and Region-Reference
Kai Shen, Lingfei Wu, Siliang Tang, Fangli Xu, Bo Long, Yueting Zhuang, Jian Pei
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 |
|---|---|---|---|
| add visual hints prediction module (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4: 21.67 -> 22.43 | Improved |
| add visual hints prediction module (ablation) | DH-Graph2Seq · VQA2.0 val | METEOR: 21.93 -> 22.57 | Improved |
| add both visual and answer hints instead of only answer type (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4: 18.53 -> 22.43 | Improved |
| add double-hints guided graph (GNN) encoder (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4: 22.17 -> 22.43 | Improved |
| use GNN encoder instead of transformer encoder for objects (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4: 22.30 -> 22.43 | Improved |
| add visual-hint-masked attention in decoding (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4: 21.98 -> 22.43 | Improved |
| add GNN and visual attention together (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4: 21.92 -> 22.43 | Improved |
| add position and answer prediction auto-encoder (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4: 22.34 -> 22.43 | Improved |
| use transformer decoder instead of LSTM decoder | DH-Graph2Seq · VQA2.0 val | CIDEr: 180.55 -> 180.66 | No clear effect |
| increase graph sparsity threshold epsilon toward 1 (ablation) | DH-Graph2Seq · VQA2.0 val | BLEU@4 | Hurt |
| add VQG-generated questions (Radial) as VQA data augmentation | BUTD/BAN/MCAN · VQA2.0 | top-1 accuracy | Hurt |
| add DH-Graph2Seq generated questions as augmentation | BUTD/BAN/MCAN · VQA2.0 | top-1 accuracy | Improved |
Closest papers by meaning
- Learning to Generate Visual Questions with Noisy Supervision (NeurIPS 2021 · similarity 0.95)
- Explicitly Guided Difficulty-Controllable Visual Question Generation (AAAI 2025 · similarity 0.91)
- ConVQG: Contrastive Visual Question Generation with Multimodal Guidance (AAAI 2024 · similarity 0.91)
- VQ-VA World: Towards High-Quality Visual Question-Visual Answering (CVPR 2026 · similarity 0.91)
- Q&A Prompts: Discovering Rich Visual Clues through Mining Question-Answer Prompts for VQA requiring Diverse World Knowledge (ECCV 2024 · similarity 0.90)
- Information Maximizing Visual Question Generation (CVPR 2019 · similarity 0.90)
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.
