RaBitQ — Quantizing High-Dimensional Vectors with Theoretical Error Bound for ANNS

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.

Research paper by Chen et al. examining the meaningfulness of nearest neighbor search in high-dimensional spaces. Analyzes limitations and implications for vector similarity search. Key for understanding ANN effectiveness.

Framework by Wei et al. for high-dimensional ANN search using subspace collision techniques. Offers efficiency and accuracy improvements for vector databases.

SIGKDD 2025 paper proposing a progressive distribution transformation approach for boosting deep vector quantization, improving quantization accuracy by progressively adapting data distributions during training.

VLDB 2024 paper presenting DIDS, a fast similarity search method using double indices and double summarizations to accelerate high-dimensional vector queries.

Research paper on high-dimensional approximate nearest neighbor search focusing on reliable and efficient distance comparison operations. Published in Proceedings of the ACM on Management of Data, Volume 1, Issue 2 in 2023 by Jianyang Gao and Cheng Long.