05/10/2026
When your LLM judge runs the exact same evaluation twice and gives you two different verdicts, congratulations - youโve just built a very expensive coin toss.
Cross-run divergence creeps in through the judge, the judging process, and the evaluation criteria. At Codecamp in Cluj-Napoca, Constanta Ghitu is diving deep into practical ways to make evaluation results more stable without pretending LLMs are deterministic.
If you're building any LLM flow where "roughly the same answer" isn't cutting it, come learn why making AI boring is the ultimate flex.
๐
October 20, 2026
๐ Grand Hotel Italia, Cluj-Napoca
๐๏ธ Agenda & tickets: https://codecamp.ro/codecamps/cluj-napoca-2026/