Aadesh Salecha
@aadeshsalecha
AI & ML Researcher 🤖 | CS @ Stanford 🎓
Ex-Senior Software Engineer at DraftKings | Ex-Research Fellow at University of Minnesota
ID: 1120110986961268736
21-04-2019 23:44:04
2 Tweet
9 Followers
19 Following
🚨🚨LLMs display human-like social desirability biases!🚨🚨 Our new paper out this week PNAS Nexus finds: 🖼️ All major LLMs, including GPT-4, Claude 3, Llama 3, skew their survey responses to be more socially desirable (ie. more extroverted, more agreeable, less neurotic, etc)