English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

DSLE: A Learning Environment Benchmarking RL Agents on Dark Souls Boss Encounters

Forum topic · 小凯 · 2026-08-11

Summary

Researchers introduce the Dark Souls Learning Environment (DSLE), a containerized platform exposing all 22 boss encounters of Dark Souls: Remastered as game-playing agent benchmarks through a Gymnasium-style interface. DSLE combines real-time combat, high-dimensional visual input, and sparse terminal rewards, where every environment step is a real action executed against the running game. The paper defines DSLE-5, a representative five-boss subset covering melee combat, a spatially constrained arena, environmental hazards, multi-target fights, and a fast final boss, proposed as a standard starting suite. Evaluations on DSLE-5 compare 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%, but none of the five methods beat the other four bosses. PPO and DQN show no measurable learning within budgets of tens of hours per run (at most 0.33% win rate on the tutorial boss, 0% elsewhere). An extended study across all 22 encounters shows the evolutionary baseline wins only a few additional early-game bosses even with fully maxed stats. Failures, from sub-10-second deaths to minute-long stalemates, are reported via survival time and damage dealt as well as win rate.

Overview

  • Field: Machine Learning
  • Authors: Derin Gezgin, Jim O'Connor, Tanner Goodwin
  • Published: 2026-08-11
  • arXiv: 2508.03798
  • Abstract

    We introduce 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. 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
  • A fast final-boss fight
  • DSLE-5 is recommended as the starting suite for agents built on DSLE.

    Evaluation Results

    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 the evolutionary baseline each defeated the game's tutorial boss, the Asylum Demon, with peak win rates of 63% and 43%, respectively.
  • None of the five methods could defeat the other four bosses in DSLE-5.
  • PPO and DQN showed no measurable learning within budgets that already required tens of hours of wall-clock time per run (at most 0.33% win rate on the tutorial boss; 0% elsewhere).
A broader study ran the evolutionary baseline across all 22 encounters; even with all 50 levels of investment in every stat, it won only a handful of additional early-game bosses, failing everywhere else.

Failure Analysis

Failure cases range from deaths in under 10 seconds in narrow, multi-target encounters to stalemates lasting nearly a minute with almost no damage dealt. The authors therefore report results via survival time and damage dealt, in addition to win rate.

---

*Auto-collected on 2026-08-12.*

Tags

#machine-learning#reinforcement-learning#dark-souls#game-benchmarks#ppo#dqn#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178633342