Research map / Large language models / Agents, code and math
Enhancing Robot Program Synthesis Through Environmental Context
Tianyi Chen, Qidi Wang, Zhen Dong, Liwei Shen, Xin Peng
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 partial environmental observations and code symbol alignment (ablation) | EVAPS program synthesis · VizDoom Top-1, Top-5, and Top-20 | Generalization Match: 30.29% (5.36%) -> 48.90% (2.23%); 49.60% (1.50%) -> 67.82% (1.81%); 61.96% (1.57%) -> 76.84% (1.13%) | Improved |
| add code symbol alignment to partial environmental observations (ablation) | EVAPS+O · VizDoom Top-1 | Generalization Match: 45.24% (2.49%) -> 48.90% (2.23%) | Improved |
| add code symbol alignment to partial environmental observations (ablation) | EVAPS+O · VizDoom Top-5 | Generalization Match: 59.71% (1.72%) -> 67.82% (1.81%) | Improved |
| add code symbol alignment to partial environmental observations (ablation) | EVAPS+O · VizDoom Top-20 | Generalization Match: 69.63% (2.22%) -> 76.84% (1.13%) | Improved |
| add partial environmental observations to code symbol alignment (ablation) | EVAPS+S · VizDoom Top-1 | Generalization Match: 41.09% (2.79%) -> 48.90% (2.23%) | Improved |
| add partial environmental observations to code symbol alignment (ablation) | EVAPS+S · VizDoom Top-5 | Generalization Match: 56.58% (3.20%) -> 67.82% (1.81%) | Improved |
| add partial environmental observations to code symbol alignment (ablation) | EVAPS+S · VizDoom Top-20 | Generalization Match: 67.78% (2.56%) -> 76.84% (1.13%) | Improved |
| use partial environmental observations without code symbol alignment (ablation) | EVAPS · VizDoom Top-1, Top-5, and Top-20 | Generalization Match | Improved |
| use code symbol alignment without partial environmental observations (ablation) | EVAPS · VizDoom Top-1, Top-5, and Top-20 | Generalization Match | Improved |
| introduce noise into observed environment grids | EVAPS · VizDoom noisy observations | Generalization Match: about 7.5%; 12% | Hurt |
| increase task complexity | EVAPS and compared methods · VizDoom tasks requiring 2, 3, 4, 5, 6, or 7+ steps | Generalization Match | Hurt |
Closest papers by meaning
- Leveraging Grammar and Reinforcement Learning for Neural Program Synthesis (ICLR 2018 · similarity 0.89)
- Towards Synthesizing Complex Programs From Input-Output Examples (ICLR 2018 · similarity 0.87)
- Write, Execute, Assess: Program Synthesis with a REPL (NeurIPS 2019 · similarity 0.87)
- HYSYNTH: Context-Free LLM Approximation for Guiding Program Synthesis (NeurIPS 2024 · similarity 0.87)
- Synthesizing Environment-Aware Activities via Activity Sketches (CVPR 2019 · similarity 0.87)
- Program Synthesis via Test-Time Transduction (NeurIPS 2025 · similarity 0.86)
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.
