Research map / Multi-agent systems and games / Games and equilibria
Evolving Action Abstractions for Real-Time Planning in Extensive-Form Games
Julian R. H. Mariño, Rubens O. Moraes, Claudio Toledo, Levi H. S. Lelis
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 evolved action abstraction (PGS*) instead of rush abstraction PGSr | PGS · uRTS 24x24 direct match | win rate (%): PGSr wins 0.4% of matches vs PGS* | Improved |
| use evolved abstraction (SSS*) instead of SCV-strategy abstraction SSSs | SSS · uRTS 24x24 direct match | win rate (%): SSS* wins 93.2% vs SSSs | Improved |
| use evolved abstractions vs hand-crafted abstraction for win rate against WR | PGS/SSS · uRTS 24x24 vs WR | win rate (%): 90.3 -> 78.8 (PGSs vs PGS*) | Hurt |
| evolved abstraction vs STT (2017 competition winner) | PGS/SSS · uRTS 24x24 | win rate (%): 99.8 (PGS*), 98.6 (SSS*) | Improved |
| evolved abstraction vs STT | PGS/SSS · uRTS 64x64 | win rate (%): 95.2 (PGS*), 91.0 (SSS*) | Improved |
| evolved abstraction vs SCV | PGS/SSS · uRTS 32x32 | win rate (%): 86.5 (PGS*), 91.5 (SSS*) | Improved |
| evolved abstraction vs PS | PGS/SSS · uRTS 64x64 | win rate (%): 83.8 (PGS*), 77.5 (SSS*) | Improved |
| use PGS* instead of SSS* as search algorithm | PGS/SSS · uRTS 64x64 head to head | win rate (%): SSS* wins 62.8% vs PGS* | No clear effect |
| evolve subset from 300 generated strategies instead of 8 hand-designed (win vs baselines grows with map size) | PGS/SSS · uRTS maps | win rate (%): 24x24 to 64x64 | Improved |
| evolutionary subset selection converges to WR-dominated population (ablation) | evolutionary algorithm · uRTS 8x8, |Z|=8,10,15 | WR fraction in population | Improved |
Closest papers by meaning
- PathWise: Planning through World Model for Automated Heuristic Design via Self-Evolving LLMs (ICML 2026 · similarity 0.86)
- TodoEvolve: Learning to Architect Agent Planning Systems (ICML 2026 · similarity 0.86)
- The Update-Equivalence Framework for Decision-Time Planning (ICLR 2024 · similarity 0.86)
- Abstraction of Situation Calculus Concurrent Game Structures (AAAI 2024 · similarity 0.86)
- Multiagent Gumbel MuZero: Efficient Planning in Combinatorial Action Spaces (AAAI 2024 · similarity 0.85)
- Strategic Planning: A Top-Down Approach to Option Generation (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.
