Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate

About

Vector quantization, a problem rooted in Shannon's source coding theory, aims to quantize high-dimensional Euclidean vectors while minimizing distortion in their geometric structure. We propose TurboQuant to address both mean-squared error (MSE) and inner product distortion, overcoming limitations of existing methods that fail to achieve optimal distortion rates. Our data-oblivious algorithms, suitable for online applications, achieve near-optimal distortion rates (within a small constant factor) across all bit-widths and dimensions. TurboQuant achieves this by randomly rotating input vectors, inducing a concentrated Beta distribution on coordinates, and leveraging the near-independence property of distinct coordinates in high dimensions to simply apply optimal scalar quantizers per each coordinate. Recognizing that MSE-optimal quantizers introduce bias in inner product estimation, we propose a two-stage approach: applying an MSE quantizer followed by a 1-bit Quantized JL (QJL) transform on the residual, resulting in an unbiased inner product quantizer. We also provide a formal proof of the information-theoretic lower bounds on best achievable distortion rate by any vector quantizer, demonstrating that TurboQuant closely matches these bounds, differing only by a small constant ($\approx 2.7$) factor. Experimental results validate our theoretical findings, showing that for KV cache quantization, we achieve absolute quality neutrality with 3.5 bits per channel and marginal quality degradation with 2.5 bits per channel. Furthermore, in nearest neighbor search tasks, our method outperforms existing product quantization techniques in recall while reducing indexing time to virtually zero.

Amir Zandieh, Majid Daliri, Majid Hadian, Vahab Mirrokni• 2025

Related benchmarks

TaskDatasetResultRank
Language ModelingWikiText-2
Perplexity (PPL)10.34
2320
Language ModelingWikiText-103 (test)
Perplexity9.22
703
Code GenerationMBPP
Pass@159.8
211
Code GenerationHumanEval
pass@163.42
145
Long-context Language UnderstandingLongBench-e
Average Score43.2
93
Mathematical ReasoningMinerva MATH500
Score21.2
36
Mathematical ReasoningGSM8K
Flexible Extract Score48.9
35
Video GenerationCausVid
LPIPS0.045
30
Code GenerationLiveCodeBench v6
Accuracy58.48
24
Mathematical Problem SolvingMATH 500
Accuracy95.39
24
Showing 10 of 31 rows

Other info

Follow for update