File size: 1,058 Bytes
2e73969 8ee7de1 dc88e16 8ee7de1 63ce9b1 8ee7de1 735be4f 8ee7de1 dc88e16 8ee7de1 dc88e16 8ee7de1 dc88e16 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
---
title: README
emoji: 🚀
colorFrom: pink
colorTo: green
sdk: static
pinned: false
---
# 🪄 GraphWiz: An Instruction-Following Language Model for Graph Problems
Project Page: [https://graph-wiz.github.io/](https://graph-wiz.github.io/)
Paper: [https://arxiv.org/abs/2402.16029.pdf](https://arxiv.org/abs/2402.16029)
Code: [https://github.com/nuochenpku/Graph-Reasoning-LLM](https://github.com/nuochenpku/Graph-Reasoning-LLM)
## About Mathoctopus
This project aims at leveraging instruction-tuning to build a powerful instruction-following LLM that can map textural descriptions of graphs and structures, and then solve different graph problems explicitly in natural language.
- **GraphWiz**, a series of instruction-following LLMs that have strong graph problem-solving abilities and output explicit reasoning paths.
- **GraphInstruct**, which offers over 72.5k training samples across nine graph problem
tasks, ranging in complexity from linear and polynomial to NP-complete, extending the scope, scale,
and diversity of previous benchmarks.
|