Sap Receives Amazon Research Award for Generative AI Work

LTI faculty member honored for his work on preventing large language models from producing factually incorrect, unethical or toxic content

Maarten Sap, an assistant professor in the Language Technologies Institute, received an Amazon Research Award for his work on preventing large language models from producing factually incorrect, unethical or toxic content.

Sap's proposal, "RLKF: Mitigating Factual Hallucinations and Social Biases With Knowledge-Based Reinforcement Learning," was one of nine selected by Amazon. RLKF stands for reinforcement learning through knowledge feedback, a play on reinforcement learning from human feedback, which powered ChatGPT's impressive abilities.

Amazon focused its spring 2023 research awards on generative artificial intelligence, reflecting the company's ongoing efforts to collaborate with researchers in this important field. Interest was high. Amazon received more submissions to its generative AI call for proposals than any other topic since the awards started in 2015.  

Sap will have access to more than 300 Amazon public datasets and can use Amazon Web Services AI and machine learning resources and tools. He can also work with Amazon researchers and participate in events and training sessions with the company.

More information about the recent Amazon Research Awards is available on the Amazon Science website.