Good Papers
NeurIPS 2026Efficient inference & servingU Wisconsin - MadisonKoreaU Wisconsin-Madison

NPUsper: Eliminating Redundant Computation for Real-Time Whisper on Mobile NPUs

NPUsper eliminates redundant Whisper computation on mobile NPUs via online hallucination detection and chunked decoding to cut latency, TTFT, and power.

Hojeong Lee, Si H Lee, Sungwon Woo, Chengpo Yan, Suman Banerjee, Seyeon Kim

Published 2026Atlanta Poster Session 5 · Fri, Dec 11, 10:00 AM–1:00 PM local time · Hall C1arXiv ↗OpenReview ↗

78%
OverallHighly rated
?
OverallHighly ratedVote to see the scoreThe exact score shows once you've voted, so every vote is your own call. The first half of each home page shelf shows its scores.
Readers
–

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel11/20reviewers recommend it
lenient 5/5
medium 6/10
strict 0/5
AI panel?Vote to see what the 20 AI reviewers said

Abstract

We present NPUsper, a live transcription system that makes Whisper efficient on mobile NPUs by eliminating redundant computation. To avoid the heavy padding used by prior streaming systems, NPUsper detects hallucinated tokens online from temporal patterns in decoder cross-attention, allowing each inference round to process short audio inputs with minimal carryover. For efficient mobile-NPU execution, we propose controlled unrolling, which executes autoregressive decoding as K-step chunk graphs, removing unnecessary KV-cache computation and reducing graph-dispatch overhead. NPUsper achieves up to 4.84x lower per-word latency, up to 33.2x lower time-to-first-token (TTFT), and up to 88.64% lower average power consumption compared with baselines, while maintaining comparable transcription accuracy. The code is available at https://github.com/npusper/NPUsper.