{BehaviorFormer}: Discovering Behavioral Patterns via Clusterable Next-Token Prediction
Discovering behavioral patterns in open-ended games is challenging due to the lack of clear objectives or ground-truth annotations. In this work, we extend a next-token prediction Transformer with the deep clustering algorithm DCSS to learn meaningful representations with well-separated clusters. In a controlled Minecraft session with 26 players, the method discovers 23 distinct behavioral patterns that generalize across the studied players, further refining structures present in self-supervised representations into clearly separable clusters. We group the clusters into seven interpretable categories covering stationary, roaming, travel, mining, base work, logging, and farming behavior.
- Veröffentlicht in:
2026 IEEE Conference on Games (CoG) - Typ:
Inproceedings - Autoren:
- Jahr:
2026 - Source:
https://arxiv.org/abs/2610.16322
Informationen zur Zitierung
: {BehaviorFormer}: Discovering Behavioral Patterns via Clusterable Next-Token Prediction, 2026 IEEE Conference on Games (CoG), 2026, https://arxiv.org/abs/2610.16322, Schneider.etal.2026a,
@Inproceedings{Schneider.etal.2026a,
author={Schneider, Tobias; Fahrner, Danial; Antebi, Soren; Uedelhoven, Daniel; Bodden, Lennard; Halscheidt, Sandra; Sifa, Rafet},
title={{BehaviorFormer}: Discovering Behavioral Patterns via Clusterable Next-Token Prediction},
booktitle={2026 IEEE Conference on Games (CoG)},
url={https://arxiv.org/abs/2610.16322},
year={2026},
abstract={Discovering behavioral patterns in open-ended games is challenging due to the lack of clear objectives or ground-truth annotations. In this work, we extend a next-token prediction Transformer with the deep clustering algorithm DCSS to learn meaningful representations with well-separated clusters. In a controlled Minecraft session with 26 players, the method discovers 23 distinct behavioral...}}
