Naive Prompt Optimization: Rethinking the Need for Complex Prompt Search
Yuan Chang and Xiaoqi Chen show that a single-lineage prompt optimizer with rollout feedback matches or beats GEPA using fewer rollouts, and that the gap widens as the teacher model gets stronger.
Ask this paper
NPO is deliberately unsophisticated: One lineage, iterative revision, a teacher model reading rollout feedback. No population, no Pareto frontier, no reflective tree search. That is the whole method.
Teacher reasoning substitutes for search: The advantage over GEPA grows with stronger teachers, which is the paper's real claim: optimizer-side search complexity is partly compensating for weak teacher reasoning, and it stops paying once the teacher is good enough.
Fewer rollouts for equal or better results: Rollout count is the cost that matters in prompt optimization, and NPO wins on it while staying broadly competitive in interactive game settings.
Why it matters: Prompt optimization is drifting toward elaborate search machinery. This is a well-argued baseline check, and the teacher-strength interaction is the kind of result that should change what people build.
Abstract
Efficiently improving autonomous agents across diverse tasks is central to accelerating recursive self-improvement (RSI) in agentic AI, with prompt optimization emerging as a promising approach capable of delivering performance gains comparable to those achieved by fine-tuning model weights, while reducing computational costs in both optimization and serving. However, recent developments increasingly favor unnecessarily complex prompt optimizers. We introduce Naive Prompt Optimization (NPO), a lightweight single-lineage method that iteratively revises prompts using a teacher model with rollout feedback. NPO achieves comparable or better performance than GEPA with fewer rollouts, and its advantage increases with stronger teacher models, suggesting that stronger teacher reasoning can partially substitute for optimizer-side search complexity. In interactive games, NPO remains broadly competitive with GEPA, while GRPO performs better on some tasks less amenable to prompt optimization. We also show that NPO-optimized prompts elicit similar performance improvements when applied verbatim to other student models, especially across models within the same family. Overall, our preliminary results show that simple, linear prompt optimization can rival substantially more sophisticated and complex search procedures.