Research map / Multi-agent systems and games / Cooperative multi-agent RL
TABX: A High-Throughput Sandbox Battle Simulator for Multi-Agent Reinforcement Learning
Hayeong Lee, JunHyeok Oh, Byung-Jun Lee
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 centralized value learning (MAPPO) instead of independent learning (IPPO) in scenarios needing global information (clover, ribbon) | MAPPO · TABX clover/ribbon | value estimation error | Improved |
| use centralized critic instead of independent learning where local observations suffice (crossfire, vsrangers, ambush) | MAPPO · TABX crossfire/vsrangers/ambush | win rate | Hurt |
| QMIX monotonic value decomposition vs IQL | QMIX · TABX scenarios | win rate | Improved |
| add RND intrinsic reward on global state to MAPPO (ablation) | MAPPO · TABX pingpong and encirclement | win rate | Improved |
| raise entropy coefficient to 0.01 in MAPPO (ablation) | MAPPO · TABX encirclement | win rate | Improved |
| replay-based UED (PLR, ACCEL, SFL) instead of domain randomization when varying unit specifications | MAPPO · TABX zero-shot unit specifications | zero-shot win rate | Hurt |
| train with health variation only vs speed or attack damage variation only (ablation) | UED baselines · TABX zero-shot generalization | zero-shot win rate | Hurt |
| UED methods on unit-specification generalization | UED baselines · TABX unseen unit specifications | average win rate: below 50% | No clear effect |
| recompilation-free reconfiguration in TABX vs SMAX | TABX · 100 random scenario resets | environment steps per second | Improved |
| TABX vs SMAX in fixed configuration | TABX · fixed configuration throughput | environment steps per second | Hurt |
Closest papers by meaning
- FightLadder: A Benchmark for Competitive Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.89)
- JaxMARL: Multi-Agent RL Environments and Algorithms in JAX (NeurIPS 2024 · similarity 0.89)
- Jumanji: a Diverse Suite of Scalable Reinforcement Learning Environments in JAX (ICLR 2024 · similarity 0.87)
- Craftax: A Lightning-Fast Benchmark for Open-Ended Reinforcement Learning (ICML 2024 · similarity 0.86)
- DesignX: Human-Competitive Algorithm Designer for Black-Box Optimization (NeurIPS 2025 · similarity 0.85)
- NAVIX: Scaling MiniGrid Environments with JAX (NeurIPS 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.
