LLM4RecAgent

A research project on recommendation system intelligent agents based on Large Language Models.

Paper Reproduction

We reproduce the paper "Can Language Models Solve Graph Problems in Natural Language?" (NeurIPS 2023 Spotlight).

The original implementation can be found at Arthur-Heng/NLGraph.

Runing Instruction

git clone --recursive https://github.com/Arlene036/LLM4RecAgent.git
pip install -r requirements.txt
export OPENAI_API_KEY=...
cd baselines
bash eval.sh

Inference from "3 Columns data"

1. Prepare your "3 columns dataset", which should be a csv file, STRICTLY COLUMN NAMES: user_id, groundtruth, top10_recommendations.

2. run, change the config in it

   bash construct.sh
   

3. Go to Colab, connect to T4 GPU;

   https://colab.research.google.com/drive/1iXGHxfmJSNDD3Z39f08oRTJyhEvy2s3u?usp=sharing

4. upload the generated csv file to the colab, and replace the following code

   test_data = pd.read_csv('/content/natural_language_top10_sample.csv')