Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

DSLE: A Learning Environment for Dark Souls Boss Encounters

Source

arXiv

arxiv.org/abs/2608.09902v1

Read the source

Source

  • Source name: arXiv
  • Source kind: paper
  • Authors / organization: Derin Gezgin, Jim O'Connor, Tanner Goodwin, Gary B. Parker
  • Venue: arXiv: cs.AI
  • Published: 2026-08-10
  • Collected: 2026-08-11
  • URL: http://arxiv.org/abs/2608.09902v1
  • PDF: https://arxiv.org/pdf/2608.09902v1
  • DOI: UNKNOWN
  • Registry group: academic_apis_and_preprints
  • Evidence role: canonical_academic_index
  • Language: en
  • Country / region: UNKNOWN

Plain-Language Brief

The authors introduce the Dark Souls Learning Environment (DSLE), a platform for training AI agents to play boss encounters in the video game 'Dark Souls: Remastered'. The DSLE is designed as a benchmark for game-playing agents and offers 22 different boss encounters. The authors evaluate several methods for training these agents, including a random policy, an expert system, an evolutionary baseline, PPO (Proximal Policy Optimization) and DQN (Deep Q-Network), but find that none of the methods can consistently defeat more than one or two bosses.

What the Document Says

  • The DSLE is a platform for training game-playing agents on Dark Souls: Remastered
  • It consists of 22 different boss encounters
  • The authors evaluate several methods for training these agents, including a random policy, an expert system, an evolutionary baseline, PPO and DQN
  • None of the evaluated methods can defeat more than one or two bosses consistently

Key Concepts Mentioned

  • Dark Souls: Remastered
  • AI agents
  • Game-playing benchmark
  • PPO
  • DQN

Claims Made by the Document

  • None of the evaluated methods can defeat more than one or two bosses consistently

Evidence Used

  • The authors report that none of the five methods defeat more than four DSLE-5 bosses
  • A broader study running the evolutionary baseline across all 22 encounters yields wins on only a handful of additional early-game bosses

Methods / Approach

  • Random policy, Expert system, Evolutionary baseline, PPO, DQN

Limitations Stated by the Authors

  • The authors note that their evaluation is limited to the methods they have tested and may not represent all possible methods for training agents on DSLE

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The agent may struggle with cramped, multi-target encounters
  • The agent may also encounter minute-long stalemates that inflict almost no damage

Open Questions

  • What other methods could potentially be effective for training agents on DSLE?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • It is unclear if there are any other methods besides the ones tested by the authors that could be effective for training agents on DSLE
  • It may be worth exploring other approaches to training game-playing agents on challenging environments like DSLE

Raw Abstract / Excerpt

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, we 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, that we recommend as the starting suite for agents built on DSLE. On DSLE-5 we evaluate 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 defeat the Asylum Demon, the game's tutorial boss (63% and 43% peak win rates), but none of the five methods defeats the other four DSLE-5 bosses; PPO and DQN show no measurable learning (at most 0.33% win rate on the tutorial boss, 0% elsewhere) within a budget that already costs tens of wall-clock hours per run. A broader study running the evolutionary baseline across all 22 encounters under advantaged all level-50 stats yields wins on only a handful of additional early-game bosses and leaves the rest unwon. The failure cases range from sub-10-second deaths in cramped, multi-target encounters to minute-long stalemates that inflict almost no damage, and we report them through survival time and damage dealt rather than win rate alone.