Meta AI’s Shepherd Criticize Language Model Outputs to Crash Hallucinations

In a new paper Shepherd: A Critic for Language Model Generation, a Meta AI research team presents Shepherd, a language model that are explicitly tuned to critique model generated outputs as well as to generate feedbacks to suggest improvements on solving the factuality, logical errors, coherence, and alignment issues.

ai ai research alignment artificial intelligence deep-neural-networks errors generated gpt hallucinations language language model machine learning machine learning & data science meta meta ai meta ai research ml paper research research team team technology