AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Depth through Recurrence: Towards Ultra-Efficient On-Device ASR

Chen Feng, Tianyi Xu, Yicheng Lin, Jay Zhuo, Ramchalam Kinattinkara Ramakrishnan, Zhaocong Yuan, Chenzheng Su, Xiaopeng Zhang

NeurIPS 2026 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes using recurrence to reuse shared parameters for depth in an ultra-efficient on-device ASR model, aiming for high representational power at low memory and compute.

Problem

Large ASR models achieve high accuracy through depth and capacity but are infeasible on resource-constrained edge devices such as watches and glasses.

Approach

Achieve representational depth through recurrence (repeatedly applying shared layers) rather than stacking distinct layers, to get high representational power with a small parameter and compute budget.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use CKA-guided grouping instead of random contiguous grouping for shared blocks (ablation)Edge-STT block-recurrent student · LibriSpeech clean train, K in 2-5WERImproved
increase number of shared blocks K (ablation)Edge-STT · Open ASR Leaderboard avgavg WER: 13.40 -> 8.57 (K=2 -> K=10)Improved
increase number of shared blocks K (ablation)Edge-Parakeet · Open ASR Leaderboard avgavg WER: 7.62 -> 6.93 (K=4 -> K=12)Improved
add MoE-LoRA personalization to block-recurrent backboneEdge-STT K=8 · VoxCeleb2 (106 speakers)test WER: 12.19 -> 8.03Improved
increase LoRA rank from 16 to 32 then 64 (ablation)MoE-LoRA on Edge-STT K=8 · VoxCeleb2test WER: 8.18 -> 8.03 -> 8.04No clear effect
increase LoRA rank to 64 (ablation)MoE-LoRA on Edge-Parakeet K=4 · VoxCeleb2test WER: 6.07 -> 6.31 (rank 32 -> 64)Hurt
add user-embedding-conditioned routing (MoE-LoRA) versus plain LoRA at same budget (ablation)Edge-STT K=8 · VoxCeleb2avg WER: 8.69 -> 8.03Improved
add task loss to decoder KL distillation (ablation)Edge-STT K=4 · LibriSpeech clean/otheravg WER: 8.15 -> 7.02Improved
add encoder-level distillation at group boundaries (ablation)Edge-STT K=4 · LibriSpeech clean/otheravg WER: 7.02 -> 6.81Improved
use block recurrence (weight sharing) in place of independent layersSTT-480ms encoder · GPU scalar kernel latencylatency: 6.22 -> 5.13 msImproved
use block recurrence in place of independent layersSTT-480ms encoder · CPU latency (Threadripper, Apple M4 Pro)speedup: 1.03-1.07xImproved
compress with block recurrence (K=4) instead of RTN INT4 quantization at same memorySTT-480ms encoder · Open ASR Leaderboard avgavg WER: 12.26 -> 10.72Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy