11 papers
MCGI: Manifold-Consistent Graph Indexing for Billion-Scale Disk-Resident Vector Search
Dongfang Zhao
Graph-based Approximate Nearest Neighbor (ANN) search often suffers from performance degradation in high-dimensional spaces due to the Euclidean-Geodesic mismatch, where greedy rou…
SIVF: GPU-Resident IVF Index for Streaming Vector Search
Dongfang Zhao
GPU-accelerated Inverted File (IVF) index is one of the industry standards for large-scale vector search but relies on static VRAM layouts that hinder real-time mutability. Our ben…
ProHD: Projection-Based Hausdorff Distance Approximation
Jiuzhou Fu, Luanzheng Guo, Nathan R. Tallent +1
The Hausdorff distance (HD) is a robust measure of set dissimilarity, but computing it exactly on large, high-dimensional datasets is prohibitively expensive. We propose \textbf{Pr…
TokenBlowUp: Resolving Representational Singularities in LLM Token Spaces via Monoidal Transformations
Dongfang Zhao
Recent work has provided compelling evidence challenging the foundational manifold hypothesis for the token embedding spaces of Large Language Models (LLMs). These findings reveal…
On The Reproducibility Limitations of RAG Systems
Baiqiang Wang, Dongfang Zhao, Nathan R Tallent +1
Retrieval-Augmented Generation (RAG) is increasingly employed in generative AI-driven scientific workflows to integrate rapidly evolving scientific knowledge bases, yet its reliabi…
PIR-RAG: A System for Private Information Retrieval in Retrieval-Augmented Generation
Baiqiang Wang, Qian Lou, Mengxin Zheng +1
Retrieval-Augmented Generation (RAG) has become a foundational component of modern AI systems, yet it introduces significant privacy risks by exposing user queries to service provi…