Overview
Field: ML Authors: Derin Gezgin, Jim O'Connor, Tanner Goodwin arXiv: 2508.05142
This post introduces the Dark Souls Learning Environment (DSLE), a containerized platform that presents all 22 boss encounters of *Dark Souls: Remastered* as game-playing agent benchmarks through a Gymnasium-style interface.
Key points
- DSLE combines real-time combat, high-dimensional visual input, and sparse terminal rewards, with each environment step being a real action executed against the running game.
- To support controlled comparison, the authors define DSLE-5, a representative five-boss subset spanning a melee fight, a spatially constrained arena, an environmental-hazard fight, a multi-target fight, and a fast final-boss fight, recommended as the starting suite for agents built on DSLE.
- On DSLE-5, the paper evaluates a random policy, an expert system, an evolutionary baseline, and PPO and DQN agents trained from visual input.
- The expert system and evolutionary baseline each defeat the tutorial boss, the Asylum Demon, with peak win rates of 63% and 43% respectively, but none of the five approaches defeats the other four bosses in DSLE-5.
- PPO and DQN agents show no measurable learning within compute budgets of tens of wall-clock hours per run (max win rate 0.33% on the tutorial boss, 0% elsewhere).
- A broader study runs the evolutionary baseline across all 22 fights with all 50 leveling attributes, winning only a few additional early bosses.
- Failure cases range from dying within 10 seconds in narrow multi-target fights to near-one-minute stalemates dealing almost no damage.
- Results are reported via survival time and damage dealt rather than win rate alone.
- Paper: https://arxiv.org/abs/2508.05142