IGMBench
An evaluation benchmark for assessing the editing capabilities of interactive world models. It covers two games, Minecraft and Terraria, and includes 110 tasks with over 1,100 executable judgment criteria. Edit quality is assessed across four dimensions: executability, behavior, preservation, and visuals.
Timeline (last 90 days)
The team introduced the IGMWorld system and the IGMBench evaluation benchmark.
IGMBench covers two games, Minecraft and Terraria, contains 110 tasks, and defines more than 1.1K executable state and behavior judgment criteria.
IGMBench evaluation adopts four categories of deterministic checks: executability, behavior, preservation, and visual.
The strongest configuration solved 78.2% of tasks under the strict task-level criteria and achieved a pass rate of 94.8% under the standard-level evaluation.
All evaluated configurations scored below 50% on the joint visual pass rate.
All Facts (5)
IGMBench covers two games, Minecraft and Terraria, contains 110 tasks, and defines more than 1.1K executable state and behavior judgment criteria.
50%UnverifiedIGMBench evaluation adopts four categories of deterministic checks: executability, behavior, preservation, and visual.
50%UnverifiedThe strongest configuration solved 78.2% of tasks under the strict task-level criteria and achieved a pass rate of 94.8% under the standard-level evaluation.
50%UnverifiedAll evaluated configurations scored below 50% on the joint visual pass rate.
50%UnverifiedThe team introduced the IGMWorld system and the IGMBench evaluation benchmark.
50%