Psychological influence tactics impact LLM code generation, study finds

A new study published on arXiv explores how psychological influence tactics, commonly used in human communication, can affect the performance of large language models (LLMs) in code generation tasks. Researchers adapted eight influence tactics from the Yukl & Falbe taxonomy into prompt templates and tested them on five leading open-weight LLMs using the LiveCodeBench and SWE-bench Verified benchmarks. The findings indicate that certain prompt framings, particularly those conveying urgency, were associated with a decrease in code correctness and security. AI

IMPACT Understanding prompt framing can lead to more reliable and secure AI-generated code.

RANK_REASON Academic paper on LLM prompt engineering techniques. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• LiveCodeBench
• SWE-bench Verified
• Yukl & Falbe

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →