Induced Natural Language Rationales and Interleaved Markup Tokens Enable Extrapolation in Large Language Models Mirelle Candida Bueno author Carlos Gemmell author Jeff Dalton author Roberto Lotufo author Rodrigo Nogueira author 2022-12 text Proceedings of the 1st Workshop on Mathematical Natural Language Processing (MathNLP) Deborah Ferreira editor Marco Valentino editor Andre Freitas editor Sean Welleck editor Moritz Schubotz editor Association for Computational Linguistics Abu Dhabi, United Arab Emirates (Hybrid) conference publication bueno-etal-2022-induced 10.18653/v1/2022.mathnlp-1.3 https://aclanthology.org/2022.mathnlp-1.3/ 2022-12 17 24