Enhance ChatGPT with EmotionPrompts

LLMs don’t have human emotions. But they respond to emotional stimuli, a new study by researchers and Microsoft and other institutions finds.

The study proposes EmotionPrompt, a prompt-engineering technique that can help explore the emotional intelligence of LLMs.

The researchers find that EmotionPrompt enhances the performance of LLMs.

Key findings:

• EmotionPrompt is composed of 11 sentences that can be appended to the original prompt

• Examples include “This is very important to my career” and “Stay determined and keep moving forward”

• They include instructions that try to elicit findings from psychology studies in self-monitoring, social cognitive theory, and cognitive emotion regulation

• The study shows that EmotionPrompt improves the performance of LLMs in both deterministic and generative tasks

• The performance boost from EmotionPrompt was even more pronounced when used in conjunction with few-shot prompting

• EmotionPrompt can easily be integrated into any LLM application

• LlamaIndex has developed a template code for using EmotionPrompt with retrieval augmented generation (RAG)

Read the full article on TechTalks.

For more on AI research:

• LLMs can’t self-correct in reasoning tasks, DeepMind study finds

• Why universal adversarial attacks against LLMs are cause for concern

• ChatGPT will make the web toxic for its successors

• Are the emergent abilities of LLMs like GPT-4 a mirage?