Timezone: »

Factuality Enhanced Language Models for Open-Ended Text Generation
Nayeon Lee · Wei Ping · Peng Xu · Mostofa Patwary · Pascale N Fung · Mohammad Shoeybi · Bryan Catanzaro

Tue Nov 29 09:00 AM -- 11:00 AM (PST) @ Hall J #232

Pretrained language models (LMs) are susceptible to generate text with nonfactual information. In this work, we measure and improve the factual accuracy of large-scale LMs for open-ended text generation. We design the FactualityPrompts test set and metrics to measure the factuality of LM generations. Based on that, we study the factual accuracy of LMs with parameter sizes ranging from 126M to 530B. Interestingly, we find that larger LMs are more factual than smaller ones, although a previous study suggests that larger LMs can be less truthful in terms of misconceptions. In addition, popular sampling algorithms (e.g., top-p) in open-ended text generation can harm the factuality due to the ``uniform randomness'' introduced at every sampling step. We propose the factual-nucleus sampling algorithm that dynamically adapts the randomness to improve the factuality of generation while maintaining quality. Furthermore, we analyze the inefficiencies of the standard training method in learning correct associations between entities from factual text corpus (e.g., Wikipedia). We propose a factuality-enhanced training method that uses TopicPrefix for better awareness of facts and sentence completion as the training objective, which can vastly reduce the factual errors.

Author Information

Nayeon Lee (HKUST)
Wei Ping (Nvidia)
Peng Xu (Nvidia)
Mostofa Patwary (NVIDIA)
Pascale N Fung (Hong Kong University of Science and Technology)
Pascale N Fung

Pascale Fung (馮雁) (born 1966 in Shanghai, China) is a professor in the Department of Electronic & Computer Engineering and the Department of Computer Science & Engineering at the Hong Kong University of Science & Technology(HKUST). She is the director of the newly established, multidisciplinary Centre for AI Research (CAiRE) at HKUST. She is an elected Fellow of the Institute of Electrical and Electronics Engineers (IEEE) for her “contributions to human-machine interactions”[1] and an elected Fellow of the International Speech Communication Association for “fundamental contributions to the interdisciplinary area of spoken language human-machine interactions”.

Mohammad Shoeybi (NVIDIA)
Bryan Catanzaro (NVIDIA)

More from the Same Authors