Video World Models · 2026-10-07

PlaySuite: A Large-Scale Benchmark for Interactive Visual Intelligence

Dheeraj Varghese, Anna Vettoruzzo, Walter Simoncini, Michelle Lorena Acevedo Callejas, Mohammad Mahdi Derakhshani, Kristof Meding, Joaquin Vanschoren, Cees G. M. Snoek

arXiv:2610.07127 PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

PlaySuite is an open, large-scale benchmark for interactive visual intelligence across 5,734 open-source video games curated from PyWeek and itch.io. It uses a unified keyboard/mouse interaction layer for HPC closed-loop evaluation and a Video-LLM-as-a-judge that maps milestone evidence to 5 progress levels. Testing 14 open multimodal interactive models shows a perception-action gap, with recurring failures in spatial grounding, action execution, and self-correction.

Why it matters

The authors argue that static perception benchmarks miss a key intelligence aspect: acting competently in dynamic environments over long horizons. PlaySuite targets this by evaluating zero-shot-like generalization across thousands of heterogeneous games using a standardized closed-loop interface and milestone-based judge scoring. The results aim to quantify a perception-action gap and provide a reproducible testbed for progress from visual perception to goal-directed interaction.

Method

Build PlaySuite from an open game pool, filtering/sandboxing to ~17.5K executable environments and evaluating 5,734 games from PyWeek and itch.io.
Run closed-loop evaluation via a unified keyboard/mouse interaction layer on HPC clusters with batched inference (shared vLLM worker) and isolated game workers.
Score progress with a Video-LLM-as-a-judge: timestamped observable milestones mapped to 5 ordered progress levels (None..Completed), reporting mean Progress Score and fraction reaching at least Partial.

Limitation

The authors state that the benchmark uses a progress-based Video-LLM-as-a-judge and reports mean Progress Score across valid evaluations; this mean should not be interpreted as the fraction of a game completed.

Abstract (from arXiv)

Recent advances in multimodal foundation models yield strong performance on static perception and reasoning benchmarks, yet such evaluations largely overlook a central aspect of intelligence: acting competently in dynamic environments over extended time horizons. We introduce PlaySuite, a large-scale benchmark for evaluating interactive visual intelligence across more than 5K open-source video games curated from PyWeek and itch.io. Spanning diverse genres and engines, including Pygame, HTML5, Godot, and Unity, these independent games are largely out-of-distribution for current models, reducing the likelihood that success can be achieved by retrieving memorized walkthroughs or web-scale training artifacts. To enable scalable evaluation across heterogeneous titles, we develop a unified closed-loop interaction framework optimized for HPC clusters alongside a Video-LLM-as-a-judge protocol that maps observable gameplay milestones to standardized progress levels. We evaluate fourteen recent open models spanning vision-language models, computer-use agents, and vision-language-action models. Our results yield strong evidence of a perception-action gap: despite strong reasoning capabilities, current models struggle to make sustained progress and exhibit recurring failures in spatial grounding, action execution, and self-correction. PlaySuite provides a reproducible and extensible testbed for measuring progress from visual perception to goal-directed interaction, and a foundation for developing models that can act, adapt, and generalize in dynamic visual environments.

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-10-05 · Indexed by Pub-AI: 2026-10-07 UTC

preprint