Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Aug 13·5 min read
TEMPER: Testing Emotional Perturbation in Quantitative Reasoning
TL;DR
TEMPER is a framework that tests how emotional language affects quantitative reasoning in large language models. It reveals that emotional framing can reduce accuracy by 2-10 percentage points, but neutralizing emotions can recover most of this lost performance.
✦ Why It Matters
Consider implementing emotional neutralization techniques in your AI models to enhance quantitative reasoning accuracy.
Key Takeaways
- Explore the impact of emotional language on model performance.
- Utilize the TEMPER framework for testing emotional perturbation.
- Implement neutralization techniques to recover accuracy.
- Consider emotional context in real-world AI applications.
How It Works
TEMPER employs a controlled emotion translation framework that rewrites quantitative reasoning problems into emotional variants while preserving numerical relationships. This allows researchers to isolate the effects of emotional framing on model performance.