Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Learning to Perform Physics Experiments via Deep Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the learned interaction policy instead of randomized interactions | Which is Heavier agents · Which is Heavier | correct-label probability | Helped |
| use the learned interaction policy instead of randomized interactions | Towers agents · Towers | correct-label probability | Helped |
| use the learned interaction policy instead of randomized interactions | Towers agents · Towers | episode length | Helped |
| increase the discount factor from 0.95 to 0.99, reducing the cost of exploration | Which is Heavier agents · Which is Heavier | episode length | Helped |
| train at greater problem difficulty (lower β) | Which is Heavier agents · Which is Heavier | exploration before labeling | Helped |
| gather information for longer when the mass gap is smaller | Which is Heavier agents · Which is Heavier | episode length | Mixed |
| use fist actuation instead of direct actuation | Towers agents · Towers | learning speed | Helped |
| reduce the control time step below the training value | Towers agent trained with pixel observations and fist actuation · Towers | correct-answer probability | Hurt |
| replace learned interactions with randomized interactions | Which is Heavier agents · Which is Heavier | episode length | No clear effect |
Closest papers by meaning
- Kinetix: Investigating the Training of General Agents through Open-Ended Physics-Based Control Tasks (ICLR 2025 · similarity 0.85)
- I-PHYRE: Interactive Physical Reasoning (ICLR 2024 · similarity 0.84)
- Rapidly Adapting Policies to the Real-World via Simulation-Guided Fine-Tuning (ICLR 2025 · similarity 0.84)
- Stabilizing Reinforcement Learning in Differentiable Multiphysics Simulation (ICLR 2025 · similarity 0.84)
- Beyond-Expert Performance with Limited Demonstrations: Efficient Imitation Learning with Double Exploration (ICLR 2025 · similarity 0.84)
- Blindfolded Experts Generalize Better: Insights from Robotic Manipulation and Videogames (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.