Research map / Multi-agent systems and games / Games and equilibria
Tackling Sparse Rewards in Real-Time Games with Statistical Forward Planning Methods
Raluca D. Gaina, Simon M. Lucas, Diego Pérez-Liébana
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 |
|---|---|---|---|
| increase MCTS rollout length from 14 to 200 (budget scaled) | MCTS · GVGAI sparse-reward games | win rate (%): 6.90 (1.79) -> 30.98 (4.14) | Improved |
| increase RHEA rollout length from 14 to 200 | RHEA · GVGAI sparse-reward games | win rate (%): 25.59 (2.82) -> 37.04 (3.85) | Improved |
| increase rollout length to 200 | MCTS · GVGAI dense-reward games | win rate (%): 54.76 (1.65) -> 53.61 (1.52) | No clear effect |
| increase rollout length to 150 | RHEA · GVGAI dense-reward games | win rate (%): 58.04 (1.73) -> 62.63 (2.10) | No clear effect |
| increase MCTS rollout length to 200 on Chopper | MCTS · Chopper | win rate (%): 100% -> 4% | Hurt |
| increase RHEA rollout length to 150 on Bait | RHEA · Bait | win rate (%): 16.5% -> 37.4% | Improved |
| add dynamic rollout length (fitness flatness) with shift buffer | RHEA · GVGAI 20 games | overall win rate (%): 48.31 (2.05) -> 21.01 (2.22) | Hurt |
| add dynamic rollout length | MCTS · GVGAI sparse-reward games | win rate (%): 6.90 (1.79) -> 19.70 (3.37) | Improved |
| add dynamic rollout length | MCTS · GVGAI 20 games | overall win rate (%): 40.40 (1.69) -> 44.04 (2.33) | Improved |
| remove shift buffer from RHEA with dynamic rollouts (ablation) | RHEA-dyn · GVGAI 20 games | overall win rate (%): 21.01 -> 39.55 | Improved |
| dynamic rollouts on RHEA, Chopper game | RHEA · Chopper | win rate (%): 100% -> 56.57% | Hurt |
| dynamic rollouts on MCTS, Wait for Breakfast | MCTS · Wait for Breakfast | win rate (%): 4% -> 42.42% | Improved |
Closest papers by meaning
- Keeping Your Distance: Solving Sparse Reward Tasks Using Self-Balancing Shaped Rewards (NeurIPS 2019 · similarity 0.87)
- Exploiting Opponents Under Utility Constraints in Sequential Games (NeurIPS 2021 · similarity 0.86)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Action-Dependent Optimality-Preserving Reward Shaping (ICML 2025 · similarity 0.85)
- Online Robust Reinforcement Learning Through Monte-Carlo Planning (ICML 2025 · similarity 0.85)
- Foundations of Multi-Agent Learning in Dynamic Environments: Where Reinforcement Learning Meets Strategic Decision-Making (AAAI 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.
