ACL Anthology · ACL 2026:acl:2026.acl-long.644 · 记录 #1
Hyperion: Private Token Sampling with Homomorphic Encryption
Lawrence LimJiaming LiuVikas KalagiDivyakant Agrawal University of California, Santa BarbaraAmr El Abbadi University of California, Santa Barbara
来源日期（仅到月份） 2026-07
来源记录日期 2026-07
雷达观测 2026-11-01 22:00
摘要
A promising direction for enabling private queries to large language models (LLMs) is with homomorphic encryption (HE). An open problem is performing token sampling under HE. In this paper, we introduce Hyperion, an efficient HE algorithm for inverse transform sampling, enabling private token sampling with 1 comparison depth, O(1) amortized comparisons, and O(\log n) rotations. We implement our approach and demonstrate that it samples tokens in 0.14 seconds for 32k tokens (\approx 4.4\mu \mathrm{s} per token) on GPU, achieving a 100\times latency improvement over prior work.