Reproduce “PromptOps: Iterative Prompt Optimization for LLM-as-a-Judge Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sven Kleinkop, Tamás Janusko, Ricardo Bochnia, Gunnar Hempel, Steffen Tomschke, Jürgen Anke, Maik Thiele
The authors present PromptOps, a framework for maintaining LLM judges by using accumulated feedback to decide when and how to revise a judge prompt. In controlled experiments, reference-score distributions shift during 500-turn e-commerce chatbot streams. Recalibration restored an advantage over a static prompt, but most final results remained below pre-shift performance; these tests do not establish live-deployment outcomes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.