Research map / Large language models / Retrieval and knowledge
Neural Question Generation with Answer Pivot
Bingning Wang, Xiaochuan Wang, Ting Tao, Qi Zhang, Jingfang Xu
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 answer pivot layer (hidden answer predictor) (ablation) | Transformer question generator, answer-agnostic · SQuAD NQG test | BLEU-4: 13.09 -> 14.78 | Improved |
| add answer pivot layer (ablation) | Transformer question generator · SQuAD NQG test | Meteor: 17.24 -> 18.61 | Improved |
| add answer pivot layer (ablation) | Transformer question generator · SQuAD NQG test | Rouge-L: 40.98 -> 41.87 | Improved |
| add answer pivot layer (ablation) | Transformer question generator · Human evaluation | Average rating: 3.63 -> 3.97 | Improved |
| add supervised answer pivot loss L_S with ground-truth answers (ablation) | answer pivot model · Human evaluation | Average rating: 3.97 -> 4.14 | Improved |
| add supervised answer pivot loss L_S (ablation) | answer pivot model · Answer pivot quality | F1: 66.2 -> 83.9 | Improved |
| add question generation objective (joint model) vs answer-only sequence labeling (ablation) | answer pivot predictor · Answer pivot quality | F1: 80.5 -> 83.9 | Improved |
| add auxiliary pivot prediction loss L_AP | answer pivot model · SQuAD NQG | posterior collapse of pivot | Improved |
| answer pivot model vs answer-aware baselines (pivot predicted rather than given) | NQG · SQuAD NQG answer-aware | BLEU-4: 15.27 -> 16.42 | Improved |
| answer pivot model vs NQG++ on answerability | NQG · Human evaluation | Answerable: 4.64 -> 3.98 | Hurt |
| use generated questions from our model as additional MRC training data vs NQG++ and SynNet data | BERT base on SQuAD · SQuAD MRC | EM / F1 | Improved |
Closest papers by meaning
- Improving Neural Question Generation Using Answer Separation (AAAI 2019 · similarity 0.91)
- Generative Question Answering: Learning to Answer the Whole Question (ICLR 2019 · similarity 0.88)
- Explicitly Guided Difficulty-Controllable Visual Question Generation (AAAI 2025 · similarity 0.84)
- How Teachers Can Use Large Language Models and Bloom’s Taxonomy to Create Educational Quizzes (AAAI 2024 · similarity 0.84)
- QGen Studio: An Adaptive Question-Answer Generation, Training and Evaluation Platform (AAAI 2025 · similarity 0.83)
- Cooperative Retrieval-Augmented Generation for Question Answering: Mutual Information Exchange and Ranking by Contrasting Layers (NeurIPS 2025 · similarity 0.83)
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.
