We have some duties to do, you can pick one and leave your name behind the item;

• Hybrid embedding; (Zhihan)

  Sparse=TFIDF, Dense=OpenAI (sparse can be switched to bm25, I wrote the code for bm25 as well, just that bm25 only generates a score and needs a query as input).

• GNNs for retrieval; (Yuntong)
• Projection layer; (Yuntong)

• Convert paper's content to a pyg graph; (Zhihan)

  take an arxiv id and convert it to a graph using NER.

• Strutured prompt design;
• Generator;

• COT, ToG, RoG ...

• Introduction;
• Related Work;
• Problem Formulation;
• Experiment;
• Conclution;
• Appendix.