Author of "Tau² Benchmark: How a Prompt Rewrite Boosted GPT-5-mini by 22%"
How media typically covers Przemysław Hejman
Przemysław Hejman as author
A simple prompt rewrite using Claude boosted GPT-5-mini's success rate by 22% on the Tau² telecom benchmark, from 55% baseline to improved performance through subtle agent policy modifications.
“Author of "Tau² Benchmark: How a Prompt Rewrite Boosted GPT-5-mini by 22%"”