Research map / Learning theory and algorithms / Combinatorial and search optimization
Asymmetric Action Abstractions for Multi-Unit Control in Adversarial Real-Time Games
Rubens Moraes, Levi 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 asymmetric rather than uniform abstraction in the second search step | GAB · Zl+Dg (16 units) | win rate: 0.66 -> 0.82 | Improved |
| Use asymmetric rather than uniform abstraction in the second search step | SAB · Zl+Dg (16 units) | win rate: 0.66 -> 0.93 | Improved |
| Select unrestricted units by AV+ rather than AV- (ablation) | GAB · 20 combat configurations, N=2 | win rate against PGS: 0.69 -> 0.88 | Improved |
| Select unrestricted units by AV+ rather than AV- (ablation) | SAB · 20 combat configurations, N=2 | win rate against SSS: 0.69 -> 0.89 | Improved |
| Select unrestricted units by AV+ rather than random selection (ablation) | GAB · 20 combat configurations, N=2 | win rate against PGS: 0.78 -> 0.88 | Improved |
| Select unrestricted units by AV+ rather than random selection (ablation) | SAB · 20 combat configurations, N=2 | win rate against SSS: 0.75 -> 0.89 | Improved |
| Use the two-step PGS-then-ABCD search scheme instead of GAS | GAB · Zl, Dg, and Zl+Dg combat scenarios | win rate against GAS: 14% | Improved |
| Use GAS's single greedy asymmetric search instead of the two-step GAB scheme (ablation) | GAB · Zl, Dg, and Zl+Dg combat scenarios | win rate against PGS: 55% | Hurt |
Closest papers by meaning
- Robust and Diverse Multi-Agent Learning via Rational Policy Gradient (NeurIPS 2025 · similarity 0.86)
- Rethinking Adversarial Policies: A Generalized Attack Formulation and Provable Defense in RL (ICLR 2024 · similarity 0.85)
- The Update-Equivalence Framework for Decision-Time Planning (ICLR 2024 · similarity 0.85)
- RL-CFR: Improving Action Abstraction for Imperfect Information Extensive-Form Games with Reinforcement Learning (ICML 2024 · similarity 0.84)
- Flow Matching for Offline Reinforcement Learning with Discrete Actions (NeurIPS 2026 · similarity 0.84)
- Sequential Decision Making in Stochastic Games with Incomplete Preferences over Temporal Objectives (AAAI 2025 · similarity 0.84)
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.
