Scaling Strategy, Not Compute: A Stand-Alone, Open-Source StarCraft II Benchmark for Accessible Reinforcement Learning Research
Sourav Panda, Shreyash Kale, Tanmay Ambadkar, Abhinav Verma, Jonathan Dodge · 2026
Read Scaling Strategy, Not Compute: A Stand-Alone, Open-Source StarCraft II Benchmark for Accessible Reinforcement Learning Research by Sourav Panda, Shreyash Kale, Tanmay Ambadkar, Abhinav Verma, Jonathan Dodge (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-06608.txt.
Description
Panda and colleagues present the Two-Bridge Map Suite, an intermediate StarCraft II benchmark between full game and mini-games, where full-game space makes rewards sparse but mini-game agents saturate. By disabling resource collection, base building, fog-of-war, and other economy mechanics, the environment isolates long-range navigation and micro-combat as two tactical skills. Experiments show agents learning maneuvering and engagement without full-game costs. Two-Bridge ships as a lightweight Gym-compatible wrapper on PySC2, with maps, wrappers, and reference scripts open-sourced.
Catalog record
| CREATOR | Sourav Panda, Shreyash Kale, Tanmay Ambadkar, Abhinav Verma, Jonathan Dodge |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 48144 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.06608