Deep Space Exploration Benchmark

It aims to assess models’ domain knowledge, professional semantic understanding, and task-scenario adaptation capabilities in deep space exploration contexts. The dataset contains 10,249 evaluation samples and covers three dimensions: professional knowledge reserve, domain-specific semantic processing, and task scenario adaptation.

The evaluation tasks include factual knowledge, conceptual knowledge, procedural knowledge, named entity recognition, terminology translation, exploration method recommendation, scientific payload configuration, and payload observation window planning. This dataset can support capability evaluation, model selection, domain adaptation, and capability enhancement research for large language models in deep space exploration.