Research map / Reinforcement learning / Visual and representation RL
MathDQN: Solving Arithmetic Word Problems via Deep Reinforcement Learning
Lei Wang, Dongxiang Zhang, Lianli Gao, Jingkuan Song, Long Guo, 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 |
|---|---|---|---|
| DQN-based expression tree construction vs prior SOTA | math word problem solver · Average of AI2/IL/CC | accuracy: 64.6 -> 75.4 | Improved |
| DQN-based expression tree construction | math word problem solver · IL | accuracy: 74.8 -> 73.3 | Hurt |
| add re-order mechanism (ablation) | MathDQN · CC | accuracy: 63.7% to 75.5% | Improved |
| add re-order mechanism (ablation) | MathDQN · AI2 | accuracy: 78.5 -> 78.5 | No clear effect |
| Q-learning instead of Sarsa (ablation) | MathDQN · CC | accuracy: 71.8 -> 75.5 | Improved |
| RL framework instead of supervised operator classifier (ablation) | MathDQN · AI2 | accuracy: 76.2 -> 78.5 | Improved |
| add context feature | UnitDep · CC | accuracy: 47.3% to 53.5% | Improved |
| add context feature | UnitDep · AI2 | accuracy: 74.7% to 56.2% | Hurt |
Closest papers by meaning
- DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning (ICLR 2026 · similarity 0.88)
- QuestA: Expanding Reasoning Capacity in LLMs via Question Augmentation (ICLR 2026 · similarity 0.86)
- MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy (AAAI 2026 · similarity 0.85)
- Arithmetic Without Algorithms: Language Models Solve Math with a Bag of Heuristics (ICLR 2025 · similarity 0.85)
- Frugal LMs Trained to Invoke Symbolic Solvers Achieve Parameter-Efficient Arithmetic Reasoning (AAAI 2024 · similarity 0.85)
- DeepSearch: Overcome the Bottleneck of Reinforcement Learning with Verifiable Rewards via Monte Carlo Tree Search (ICLR 2026 · similarity 0.84)
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.
