Research map / Reinforcement learning / Policy gradient and actor-critic
Memory Augmented Policy Optimization for Program Synthesis and Semantic Parsing
Chen Liang, Mohammad Norouzi, Jonathan Berant, Quoc V Le, Ni Lao
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 |
|---|---|---|---|
| Use MAPO instead of REINFORCE | Neural Symbolic Machines · WikiTableQuestions dev | accuracy (%): < 10 -> 42.3\pm0.3 | Improved |
| Use MAPO instead of MML (Soft EM) | Neural Symbolic Machines · WikiTableQuestions dev | accuracy (%): 39.7 \pm 0.3 -> 42.3\pm0.3 | Improved |
| Use MAPO instead of REINFORCE | Neural Symbolic Machines · WikiSQL dev | accuracy (%): < 10 -> 72.2\pm0.2 | Improved |
| Use MAPO instead of MML (Soft EM) | Neural Symbolic Machines · WikiSQL dev | accuracy (%): 70.7 \pm 0.1 -> 72.2\pm0.2 | Improved |
| Add systematic exploration (ablation) | MAPO · WikiTableQuestions dev | accuracy (%): < 10 -> 42.3\pm0.3 | Improved |
| Add systematic exploration (ablation) | MAPO · WikiSQL dev | accuracy (%): < 10 -> 72.2\pm0.2 | Improved |
| Add memory weight clipping (ablation) | MAPO · WikiTableQuestions dev | accuracy (%): < 10 -> 42.3\pm0.3 | Improved |
| Add memory weight clipping (ablation) | MAPO · WikiSQL dev | accuracy (%): < 10 -> 72.2\pm0.2 | Improved |
| Use MAPO with an ensemble of 10 | MAPO · WikiTableQuestions test | accuracy (%): 43.7 -> {\bf 46.3} | Improved |
Closest papers by meaning
- Single-stream Policy Optimization (ICLR 2026 · similarity 0.88)
- TGPO: Efficient Policy Optimization through Sequence Anchor and Information Gating (ICML 2026 · similarity 0.87)
- Learning to Foresee: Unveiling the Unlocking Efficiency of On-Policy Distillation (NeurIPS 2026 · similarity 0.86)
- HAPO: Training Language Models to Reason Concisely via History-Aware Policy Optimization (AAAI 2026 · similarity 0.85)
- On the Limits of Test-Time Compute: Sequential Reward Filtering for Better Inference (ICML 2026 · similarity 0.85)
- Optimizing Language Models for Inference Time Objectives using Reinforcement Learning (ICML 2025 · 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.
