StarBench: A Turn-Based RPG Benchmark for Agentic Multimodal Decision-Making and Information Seeking

TL;DR

StarBench is a new benchmark for evaluating vision-language models on human-like decision-making and information seeking in a turn-based RPG, highlighting current gaps and potential for agentic behaviors.

Contribution

Introduces StarBench, a comprehensive RPG benchmark for assessing multimodal decision-making and agentic information seeking in realistic game scenarios.

Findings

Current VLMs show significant perception-to-control gaps.

Information seeking improves agent success rates.

Benchmark provides standardized evaluation metrics.

Abstract

Human players do more than press buttons: they ground what they see on screen into precise keyboard-mouse actions and, when stuck, they seek information before trying again. We ask whether current vision-language models (VLMs) can do the same. Despite encouraging results under simplified control or tool scaffolds, human-like play in a real client - mapping raw screenshots to temporally coherent low-level actions while deciding when to ask for guidance - remains an open challenge. We introduce StarBench, a turn-based RPG benchmark derived from Honkai: Star Rail that targets these two human-like competencies: multimodal decision-making from pixels to actions and agentic information seeking. StarBench standardizes evaluation across eight combat tasks and two regimes with shared tasks and metrics: (i) direct control, where agents receive only screenshots and must emit low-level primitives (click and keypress) with no semantic hints; and (ii) tool-assisted control, where higher-level intents can be mapped to primitives by detectors and OCR outputs provide optional textualized observations to ease UI grounding. To mirror human practice, StarBench also includes an ask-or-act diagnostic that measures whether and when agents choose to request brief guidance before proceeding, and how that choice affects subsequent performance. We report reference baselines for contemporary VLMs and a human reference. Results expose sizable gaps in perception-to-control fidelity in the direct regime, while showing that judicious information seeking correlates with improved success, establishing StarBench as a reproducible yardstick for agentic information seeking and multimodal decision-making in real-client play.