flan-t5-large-gsm8k

GRACE: Discriminator-Guided Chain-of-Thought Reasoning

This model is a fine-tuned version of google/flan-t5-large on the GSM8K dataset. It serves as the generator for the GRACE (Guiding chain-of-thought ReAsoning with a CorrectnEss Discriminator) decoding strategy.

Model Description

GRACE is a stepwise decoding approach that steers the decoding process towards producing correct reasoning steps. It employs a step-level verifier or discriminator trained with a contrastive loss over correct and incorrect steps, which is used during decoding to score next-step candidates. This specific checkpoint is the fine-tuned generator model used in the paper's experiments for tasks like GSM8K.

• Paper: GRACE: Discriminator-Guided Chain-of-Thought Reasoning
• Repository: https://github.com/mukhal/grace

Citation

If you use this model or the GRACE approach, please consider citing the following paper:

@article{khalifa2023grace,
  title={Grace: Discriminator-guided chain-of-thought reasoning},
  author={Khalifa, Muhammad and Logeswaran, Lajanugen and Lee, Moontae and Lee, Honglak and Wang, Lu},
  journal={arXiv preprint arXiv:2305.14934},
  year={2023}
}