AdaptiveIndex — Adaptive Indexing in High-Dimensional Metric Spaces

TKDE 2024 paper presenting DIMS, a distributed indexing method for efficient similarity search across metric spaces. The approach enables parallel processing of vector similarity queries at scale.

A library for efficient similarity search and creation of nearest neighbor indices in metric spaces, implementing HNSW, SW-graph, and other ANN algorithms, widely used as a backend for vector search engines.

Multi-Vector Retrieval Algorithm that reduces multi-vector similarity search to single-vector similarity search via Fixed Dimensional Encodings. Achieves 10% improved recall with 90% lower latency compared to existing approaches.

A commutative vector quantization method for KV cache compression that reduces FP16 cache size by 87.5% with 2-bit quantization and enables 1-bit quantization, allowing LLaMA-3.1 8B to run with 128K context on a single RTX 4090 GPU.

VLDB 2025 paper proposing efficient level navigation with shortcuts for accelerating approximate nearest neighbor search in hierarchical graph indexes, improving traversal speed across multi-layer graph structures.

SIGKDD 2025 paper proposing adaptive awareness capabilities for graph-based approximate nearest neighbor search, enabling the search algorithm to dynamically adjust its strategy based on local graph characteristics and query properties.