Large language models remain remarkably brittle when confronted with reasoning tasks that require careful epistemic validation. In experiments described in a new paper titled 'Pramana: Fine-Tuning Large Language Models for Epistemic Reasoning through Navya-Nyaya,' Apple researchers demonstrated that adding irrelevant context to mathematical problems caused LLM performance to degrade by 65 percent—a stark illustration of how current models fail at systematic logical reasoning. Rather than viewing this as an intractable limitation, the research team identified a solution in an unexpected source: Navya-Nyaya, a 16th-century Indian philosophical school that developed rigorous frameworks for evaluating knowledge claims and distinguishing valid reasoning from unfounded assertions. The core insight is that classical logical systems offer structured methodologies for validating inferences that modern neural networks have never explicitly learned.