Abstract: In this work, we aim to reproduce the EMNLP 2023 paper of Liu et al. (2023) titled Plan, Verify and Switch: Integrated Reasoning with Diverse X-of-Thoughts. In the original paper, XoT - a prompting method which is designed to enhance LLMs' ability to deal with mathematical problems - is introduced and tested using GPT-3.5. The experiments showed that the technique outperforms the existing ones. We seek not only to ascertain the effectiveness of the method for a smaller model, the Phi-2, but also to expand the ideas of the paper by integrating metacognitive evaluation and broadening one of its modules.
Submission Length: Regular submission (no more than 12 pages of main content)
Assigned Action Editor: ~Kuldeep_S._Meel2
Submission Number: 2187
Loading