R0023/2026-03-25/Q001/S02/R02¶
Peer-reviewed EMNLP 2024 study demonstrating that personas in system prompts do not improve LLM performance.
Summary¶
| Field | Value |
|---|---|
| Title | When "A Helpful Assistant" Is Not Really Helpful: Personas in System Prompts Do Not Improve Performances of Large Language Models |
| URL | https://aclanthology.org/2024.findings-emnlp.888/ |
| Date accessed | 2026-03-25 |
| Publication date | 2024-11 |
| Author(s) | Mingqian Zheng, Jiaxin Pei, Lajanugen Logeswaran, Moontae Lee, David Jurgens |
| Publication | Findings of EMNLP 2024 |
Selection Decision¶
Included in evidence base: Yes
Rationale: Peer-reviewed at a top NLP venue. Tests 162 distinct roles across 4 LLM families on 2,410 factual questions. Directly shows persona prompting does not improve and can degrade performance.