Research map / Reinforcement learning / Visual and representation RL
Tempo Adaptation in Non-stationary Reinforcement Learning
Hyunin Lee, Yuhao Ding, Jongmin Lee, Ming Jin, Javad Lavaei, Somayeh Sojoudi
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 |
|---|---|---|---|
| Use ProST-G instead of the strongest baseline | Swimmer-v2 · speeds 1–5 | average return: 0.57 vs -0.08; 1.04 vs 0.20; 1.52 vs -0.13; 2.01 vs -0.04; 2.81 vs 0.10 | Improved |
| Use ProST-G instead of MBPO | Halfcheetah-v2 · speeds 1–4 | average return: -19.69 vs -24.89; -20.21 vs -22.19; -21.04 vs -21.65; -19.55 vs -21.40 | Improved |
| Use ProST-G instead of MBPO | Halfcheetah-v2 · speed 5 | average return: -20.50 vs -20.52 | Hurt |
| Use ProST-G instead of the strongest baseline | Hopper-v2 · speed 1 | average return: 92.77 vs 98.38 | Hurt |
| Use ProST-G instead of the strongest baseline | Hopper-v2 · speeds 2–5 | average return: 98.13 vs 99.02; 100.42 vs 98.60; 100.68 vs 98.89; 102.48 vs 99.64 | Mixed |
| Set policy training time to Δπ=4 (ablation) | ProST-G · synthetic non-stationary environment | average reward | Improved |
| Replace ARIMA with a simple average forecaster, varying window size w (ablation) | ProST-G · four environment speeds | average reward | Mixed |
| Use a proactive ARIMA forecaster instead of MBPO's reactive model (ablation) | ProST-G / MBPO · non-stationary Mujoco tasks | average reward | Improved |
| Jointly optimize data weights and the future model with W-LSE instead of selecting a fixed sliding window with SW-LSE (ablation) | ProST-T · time-elapsing MDP | dynamic regret upper bound | Improved |
Closest papers by meaning
- Forecasting in Offline Reinforcement Learning for Non-stationary Environments (NeurIPS 2025 · similarity 0.89)
- Pausing Policy Learning in Non-stationary Reinforcement Learning (ICML 2024 · similarity 0.88)
- Enabling Realtime Reinforcement Learning at Scale with Staggered Asynchronous Inference (ICLR 2025 · similarity 0.88)
- Online Reinforcement Learning in Non-Stationary Context-Driven Environments (ICLR 2025 · similarity 0.88)
- ART for Diffusion Sampling: A Reinforcement Learning Approach to Timestep Scheduling (NeurIPS 2026 · similarity 0.88)
- Reinforcement Learning with Random Time Horizons (ICML 2025 · similarity 0.87)
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.
