Released on Nov 3, 2025
- Improved clustering for topics by incorporating paper similarity based on LLM-embeddings into the bibliographic graph and joint embeddings computation
- Reuse the embeddings DB for fetching text embeddings
- Use parquet format for Faiss index storage