TY  - RPRT
TI  - GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning
AU  - Agrawal, Lakshya A. [University of California, Berkeley, CA (United States)]
AU  - Tan, Shangyin [University of California, Berkeley, CA (United States)]
AU  - Soylu, Dilara [Stanford Univ., CA (United States)]
AU  - Ziems, Noah [University of Notre Dame, IN (United States)]
AU  - Khare, Rishi [University of California, Berkeley, CA (United States)]
AU  - Opsahl-Ong, Krista [Databricks, San Francisco, CA (United States)]
AU  - Singhvi, Arnav [Stanford Univ., CA (United States); Databricks, San Francisco, CA (United States)]
AU  - Shandilya, Herumb [Stanford Univ., CA (United States)]
AU  - Ryan, Michael J. [Stanford Univ., CA (United States)]
AU  - Jiang, Meng [University of Notre Dame, IN (United States)]
AU  - Potts, Christopher [Stanford Univ., CA (United States)]
AU  - Sen, Koushik [University of California, Berkeley, CA (United States)]
AU  - Dimakis, Alexandros G. [University of California, Berkeley, CA (United States); Bespoke Labs, Santa Clara, CA (United States)]
AU  - Stoica, Ion [University of California, Berkeley, CA (United States)]
AU  - Klein, Dan [University of California, Berkeley, CA (United States)]
AU  - Zaharia, Matei [Databricks, San Francisco, CA (United States)]
AU  - Khattab, Omar [Massachusetts Inst. of Technology (MIT), Cambridge, MA (United States)]
PY  - 2025
DO  - 10.48550/arxiv.2507.19457
UR  - https://www.osti.gov/biblio/3025768
ID  - 3025768
ER  -