Original post
Omar Khattab#160
björn@bj2rn
microsoft used gepa / dspy to tune the LLM judge prompt for quality scoring. @lateinteraction stays winning. from the mai-thinking-1 report
3:40 PM · Jun 2, 2026 · 1.7K Views
microsoft used gepa / dspy to tune the LLM judge prompt for quality scoring. @lateinteraction stays winning. from the mai-thinking-1 report
microsoft used gepa / dspy to tune the LLM judge prompt for quality scoring. @lateinteraction stays winning. from the mai-thinking-1 report
Users call the paper on Microsoft's GEPA and DSPy optimization of LLM judge prompts cool because it improves data quality.