Showing cs.DBShow all
3 papers · 1 filter
cs.DB2026
Contextual Utility of Quantization Moves in Extreme Low-Bit LLMs
Wenxuan Xiao, Xu Cao
Post-training quantizers select finite code changes using reconstruction proxies or local loss approximations, but the utility of a quantization move depends on the state through w…
cs.DB2026
When Does Low-Bit Quantization Preserve the Decisions of Vector Search?
Wenxuan Xiao, Xu Cao
Low-bit quantization can achieve high recall on some vector representations and fail sharply on others, while average distortion and global rank correlation do not explain the diff…
cs.DB2026
QuIVer: Rethinking ANN Graph Topology via Training-Free Binary Quantization
Wenxuan Xiao, Zhiyou Wang, Chengcheng Li
Approximate nearest neighbor (ANN) graph indices such as HNSW and Vamana construct their edge topology in full-precision or high-fidelity quantized metric spaces, relegating binary…