Research map / Large language models / AI and society
The Gap of Semantic Parsing: A Survey on Automatic Math Word Problem Solvers
Dongxiang Zhang, Lei Wang, Luming Zhang, Bing Tian Dai, Heng Tao Shen
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 context feature to UNITDEP classifier (vs removing it) (ablation) | UNITDEP · AI2 | accuracy: 74.7 -> 56.2 | Hurt |
| add context feature to UNITDEP classifier (vs removing it) (ablation) | UNITDEP · CC | accuracy: 47.3 -> 53.5 | Improved |
| use reinforcement-learning tree construction (MathDQN) instead of exhaustive enumeration (ALGES) | MathDQN · CC | accuracy: 65 -> 75.5 | Improved |
| use deep neural solvers on large diverse dataset instead of small ones | existing equation set solvers · Dolphin18K | accuracy | Hurt |
| add seq2seq model to similarity retrieval in DNS (ablation) | DNS · ALG514 | accuracy: 16.1% seq2seq alone; DNS 70.1 equals SIM 70.1 | No clear effect |
| add copy-alignment and policy-gradient training (CASS) to seq2seq | CASS · Dolphin18K | accuracy: 21.6 -> 29.0 | Improved |
| train with RL policy gradient instead of maximum likelihood (ablation) | CASS · equation set datasets | accuracy | Improved |
| use BiLSTM or structured self-attention instead of GRU/LSTM for template classification (ablation) | DNS follow-up · equation set datasets | accuracy | No clear effect |
| add implicit supervision from external 2,000 samples to MixedSP (ablation) | MixedSP · ALG514 | accuracy | Improved |
Closest papers by meaning
- A Compact Model for Mathematics Problem Representations Distilled from BERT (AAAI 2025 · similarity 0.88)
- Template-Driven LLM-Paraphrased Framework for Tabular Math Word Problem Generation (AAAI 2025 · similarity 0.86)
- Can LLMs Solve Longer Math Word Problems Better? (ICLR 2025 · similarity 0.85)
- RealMath: A Continuous Benchmark for Evaluating Language Models on Research-Level Mathematics (NeurIPS 2025 · similarity 0.85)
- MathCoder2: Better Math Reasoning from Continued Pretraining on Model-translated Mathematical Code (ICLR 2025 · similarity 0.85)
- OpenWebMath: An Open Dataset of High-Quality Mathematical Web Text (ICLR 2024 · similarity 0.85)
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.
