Metraipolis › archive › dc_031c4e4bd4f8

Where It Breaks: Diagnosing Perception, Memory, and Decision Failures of a Vision-Language Agent in Pac-Man — arXiv:2611.06704

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.06704
License (as archived)
CC0-1.0
Archived
2026-11-05T18:22:25.032Z
SHA-256 of the archived text
bbfe615b3538c237b53f5a2d3491473ba2084545f01cc528ab56539506b94ade
Stable URL of this page
https://metraipolis.com/documento/dc_031c4e4bd4f8

How to cite

Where It Breaks: Diagnosing Perception, Memory, and Decision Failures of a Vision-Language Agent in Pac-Man — arXiv:2611.06704 — arXiv (en), https://arxiv.org/abs/2611.06704 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_031c4e4bd4f8 (Metraipolis archive dc_031c4e4bd4f8, sha256 bbfe615b3538c237b53f5a2d3491473ba2084545f01cc528ab56539506b94ade)
BibTeX
@misc{dc031c4e4bd4f8,
  title = {Where It Breaks: Diagnosing Perception, Memory, and Decision Failures of a Vision-Language Agent in Pac-Man — arXiv:2611.06704},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.06704},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_031c4e4bd4f8; sha256 bbfe615b3538c237b53f5a2d3491473ba2084545f01cc528ab56539506b94ade},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.06704. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Training an agent to play Pac-Man with only up, down, left and right actions raises a basic question: when performance is poor, what should be improved? Difficulty can arise from reading the screen or from choosing actions over a long sequence. We examine these problems in a Qwen3.5-9B-based agent through separate perception checks and primitive-policy reinforcement-learning studies. In the policy study, which uses textual boards to remove image transcription from the decision interface, completion on eight held-out mazes increased from 0% before this optimization to 80.2% at a selected checkpoint, with three extra lives. The perception checks show that accurate static screenshot reading can coexist with errors during interaction. A separate policy comparison also shows that further optimization can accompany worse action choices. Together, these findings motivate concrete checks for post-training: evaluate perception in use, distinguish training success from full-game validation, and examine decisions and later performance rather than relying on one selected score. The study provides empirical lessons for improving future agent training and evaluation.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2611.06704 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-05T18:15:49.024Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).