higgs-llama-vicuna-ep25-70b

Higgs Model Card

Github details

Please checkout the repo: https://github.com/luffycodes/Tutorbot-Spock-Phys.

Model details

Model type: Higgs is an open-source educational tutoring chatbot trained by fine-tuning LLaMA-2-70B-chat model on synthetic student-tutorbot conversations generated using specialized prompts. Higgs performs a code soliloquy (an inner monologue) in which Higgs prompts itself through a series of prompts to decide if it's next response to the student would need any math calculations. If Higgs determines that the response might require such a calculation, such as in cases to verify student calculation, it outputs python code. Then, it uses the output of the python code to frame a mathematically accurate reply to the student.

Model date: Higgs was trained between May 2023 and Sept 2023.

Organizations developing the model: The Higgs (Spock) team with members from Rice University and OpenStax.

Where to send questions or comments about the model: Shashank Sonkar ([email protected])

If you use this work, please cite: Code Soliloquies for Accurate Calculations in Large Language Models https://arxiv.org/abs/2309.12161

@misc{sonkar2023code,
      title={Code Soliloquies for Accurate Calculations in Large Language Models}, 
      author={Shashank Sonkar and MyCo Le and Xinghe Chen and Lucy Liu and Debshila Basu Mallick and Richard G. Baraniuk},
      year={2023},
      eprint={2309.12161},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}