Thank you for sharing our work!

Database-Augmented Query Representation for Information Retrieval Enhances query representations by leveraging metadata from multiple database tables. Uses a graph-based set encoding strategy to efficiently handle large volumes of unordered metadata. 📝arxiv.org/abs/2406.16013

Thanks for the shout-out ! 🙌 Excited to see Adaptive RAG + Local LLMs in TypeScript making strides. With amazing collaborators, Jinheon Baek, Sukmin Cho, Sung Ju Hwang, and my advisor.

Lossless Acceleration of Large Language Models with Hierarchical Drafting based on Temporal Locality in Speculative Decoding huggingface.co/papers/2502.05…

8. Lossless Acceleration of Large Language Models with Hierarchical Drafting based on Temporal Locality in Speculative Decoding 🔑 Keywords: Large Language Models, Inference Speed, Hierarchy Drafting, Speculative Decoding, Drafting Strategies 💡 Category: Natural Language

This paper proposes Hierarchy Drafting (HD) to address inferencing challenge. HD organizes token sources into hierarchical databases based on temporal locality. HD accesses these databases sequentially to enhance drafting accuracy and maintain consistent acceleration across

The RAG Paradox: A Black-Box Attack Exploiting Unintentional Vulnerabilities in Retrieval-Augmented Generation Systems. arxiv.org/abs/2502.20995