GeoMind: An Agentic Workflow for Lithology Classification with Reasoned Tool Invocation

arXiv:2604.21501v1 Announce Type: new Abstract: Lithology classification in well logs is a fundamental geoscience data mining task that aims to infer rock types from multi dimensional geophysical sequences. Despite recent progress, existing approaches typically formulate the problem as a static, single-step discriminative mapping. This static paradigm limits evidence-based diagnostic reasoning against geological standards, often yielding predictions that are detached from geological reality due