We are pleased to announce that this paper has been accepted by Core Journals of China: Journal of Chinese Information Processing.
Source code for our paper :
Self-Guide: Enhancing LLM Reasoning Ability via Self-Plan. CCL2024
Self-Guide:一种基于自我规划的大语言模型推理增强方法. CCL2024
If you find this work useful, please cite our paper and give us a shining star 🌟
Despite significant advancements of LLMs in NLP tasks, they still face cognitive overload issues, especially in domains requiring complex reasoning, where the models need to process and memorize vast amounts of information during the reasoning process. Therefore, it is a pressing issue to effectively reduce the cognitive load during the reasoning process of LLM to alleviate potential cognitive overload. We introduce the Self-Guide method to address the issue, which boosts LLMs' reasoning abilities by leveraging self-generated common sense knowledge and reasoning instructions. Experimental results demonstrate that our Self-Guide method outperforms baseline methods significantly on four common reasoning tasks. By combining the self-planning and reasoning abilities of LLMs, our method provides a new and effective approach to enhance the reasoning ability of LLMs.
git clone https://github.com/10-OASIS-01/Self-Guide
pip install -r requirements.txtWe also provide the full request code, you can re-request for further exploration.
First, set your own api-key in run file:
openai.api_key = 'sk-<your-api-key>'Then, run the following script:
python run.py --dataset CLUTRR --start_index 0Analyzing log files:
python build.py --dataset CLUTRREvaluate:
python evaluate.py --dataset CLUTRRComing Soon！
If you have questions, suggestions, and bug reports, please send a email to us, we will try our best to help you.