arXiv: On Turn-Level Best-of-N Sampling for Scaling LLM Agents
This paper is closely related to the work of \cite{zainullina2025guided} who used critic models for guided search at inference time.
The arXiv Coding-Agent Research page documents the arXiv update.
