Home / Models & Research / Pramana: Fine-Tuning Large Language Models for Epistemic Reasoning Through Navya-Nyaya
Models & Research Thursday, 9 April 2026 | 1 min read

Pramana: Fine-Tuning Large Language Models for Epistemic Reasoning Through Navya-Nyaya

Researchers from Apple have developed a method to fine-tune large language models for epistemic reasoning using the Navya-Nyaya framework. The approach, called Pramana, aims to address the limitations of large language models in systematic reasoning, which often result in hallucinations and unfounded claims. The researchers added irrelevant context to mathematical problems and observed a 65% degradation in performance, demonstrating the potential of the Pramana method. This work has implications for the development of more robust and reliable language models.

Original Sources

Tags

#language models #epistemic reasoning #fine-tuning
All stories