Temperature-Dependent Performance of Prompting Strategies in Extended Reasoning Large Language Models

Develop an automated sweep utility that identifies the optimal temperature-prompting pair for reasoning tasks. This prevents wasteful spend on trial-and-error manual experimentation.

Suggested repo: temp-sweep

"Find your model's reasoning sweet spot automatically."

Estimated effort: 20h