Search papers, labs, and topics across Lattice.
This paper introduces the MineExplorer benchmark to evaluate the open-world exploration capabilities of multimodal large language models (MLLMs) in Minecraft, addressing the limitations of existing benchmarks that often rely on short-horizon tasks. By filtering atomic tasks to minimize reliance on Minecraft-specific knowledge and employing a multi-agent synthesis workflow, the authors create a more reliable evaluation framework. Experiments reveal that while advanced MLLM agents excel at single-hop tasks, they struggle with longer, multi-hop tasks that require coordination of hidden prerequisites, indicating significant challenges in open-world reasoning.
MLLMs excel at single-hop tasks but falter dramatically in open-world scenarios, revealing critical gaps in their reasoning capabilities.
Multimodal large language models (MLLMs) have shown strong capabilities in perception, reasoning, and action generation. However, their ability to sustain exploration in dynamic open worlds remains unclear. Existing embodied and game-based benchmarks often compress interaction into short-horizon tasks or entangle success with domain-specific game mechanics. In this paper, we introduce MineExplorer benchmark for evaluating open-world exploration capabilities of MLLM agents in Minecraft. We first filter atomic tasks whose solutions rely heavily on Minecraft-specific knowledge to better reflect general open-world reasoning. Then we organize the benchmark around a ReAct-style capability formulation and compose atomic tasks into implicit multi-hop tasks. To further construct reliable instances, MineExplorer uses a multi-agent synthesis workflow that jointly designs task graphs, sandbox scenes, and rule-based milestone evaluators. Human evaluation shows that the multi-agent synthesis workflow produces significantly more reliable instances than a single-agent baseline. Experiments with advanced MLLM agents show that open-world exploration remains challenging, as strong models can handle many single-hop tasks but degrade sharply when hidden prerequisites must be coordinated over longer trajectories. Further analysis finds that task difficulty tracks agent completion, and larger models or thinking modes do not consistently translate into better performance. Code and dataset are available at https://github.com/Jometeorie/MineExplorer.
