# Extreme Encoder Output Frame Rate Reduction: Improving Computational Latencies of Large End-to-End Models

arXiv: https://arxiv.org/abs/2402.17184 · HTML: https://mlsanity.com/paper/2402.17184/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 28 of its 40 references that are in the map.
- First posted on arXiv 27 Feb 2024.

## Builds on (most cited first, 28 of 28)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://mlsanity.com/paper/2006.11477/) (2020)
- [Conformer: Convolution-augmented Transformer for Speech Recognition](https://mlsanity.com/paper/2005.08100/) (2020)
- [SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition](https://mlsanity.com/paper/1904.08779/) (2019)
- [Adafactor: Adaptive Learning Rates with Sublinear Memory Cost](https://mlsanity.com/paper/1804.04235/) (2018)
- [Speech Recognition with Deep Recurrent Neural Networks](https://mlsanity.com/paper/1303.5778/) (2013)
- [Listen, Attend and Spell](https://mlsanity.com/paper/1508.01211/) (2015)
- [Transformer Transducer: A Streamable Speech Recognition Model with Transformer Encoders and RNN-T Loss](https://mlsanity.com/paper/2002.02562/) (2020)
- [Streaming End-to-end Speech Recognition For Mobile Devices](https://mlsanity.com/paper/1811.06621/) (2018)
- [Google USM: Scaling Automatic Speech Recognition Beyond 100 Languages](https://mlsanity.com/paper/2303.01037/) (2023)
- [Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing](https://mlsanity.com/paper/2006.03236/) (2020)
- [Recent Advances in End-to-End Automatic Speech Recognition](https://mlsanity.com/paper/2111.01690/) (2021)
- [End-to-End Speech Recognition: A Survey](https://mlsanity.com/paper/2303.03329/) (2023)
- [Massively Multilingual ASR: 50 Languages, 1 Model, 1 Billion Parameters](https://mlsanity.com/paper/2007.03001/) (2020)
- [Hybrid Autoregressive Transducer (hat)](https://mlsanity.com/paper/2003.07705/) (2020)
- [Lingvo: a Modular and Scalable Framework for Sequence-to-Sequence Modeling](https://mlsanity.com/paper/1902.08295/) (2019)
- [A Better and Faster End-to-End Model for Streaming ASR](https://mlsanity.com/paper/2011.10798/) (2020)
- [Minimum Word Error Rate Training for Attention-based Sequence-to-Sequence Models](https://mlsanity.com/paper/1712.01818/) (2017)
- [Scaling End-to-End Models for Large-Scale Multilingual ASR](https://mlsanity.com/paper/2104.14830/) (2021)
- [Tied & Reduced RNN-T Decoder](https://mlsanity.com/paper/2109.07513/) (2021)
- [Less Is More: Improved RNN-T Decoding Using Limited Label Context and Path Merging](https://mlsanity.com/paper/2012.06749/) (2020)
- [Pseudo Label Is Better Than Human Label](https://mlsanity.com/paper/2203.12668/) (2022)
- [Improving Tail Performance of a Deliberation E2E ASR Model Using a Large Text Corpus](https://mlsanity.com/paper/2008.10491/) (2020)
- [Dissecting User-Perceived Latency of On-Device E2E Speech Recognition](https://mlsanity.com/paper/2104.02207/) (2021)
- [A Unified Cascaded Encoder ASR Model for Dynamic Model Sizes](https://mlsanity.com/paper/2204.06164/) (2022)
- [2-bit Conformer quantization for automatic speech recognition](https://mlsanity.com/paper/2305.16619/) (2023)
- [Massive End-to-end Models for Short Search Queries](https://mlsanity.com/paper/2309.12963/) (2023)

## Built on by (most cited first, 6 of 6)

- [Text Injection for Neural Contextual Biasing](https://mlsanity.com/paper/2406.02921/) (2024)
- [Synth4Kws: Synthesized Speech for User Defined Keyword Spotting in Low Resource Environments](https://mlsanity.com/paper/2407.16840/) (2024)
- [GE2E-KWS: Generalized End-to-End Training and Evaluation for Zero-shot Keyword Spotting](https://mlsanity.com/paper/2410.16647/) (2024)
- [CTC-GMM: CTC guided modality matching for fast and accurate streaming speech translation](https://mlsanity.com/paper/2410.05146/) (2024)
- [Aligner-Encoders: Self-Attention Transformers Can Be Self-Transducers](https://mlsanity.com/paper/2502.05232/) (2025)
- [ASSET: Adaptive Speech Segmentation via Reinforcement Learning](https://mlsanity.com/paper/2610.14309/) (2026)

## Abstract (arXiv)

The accuracy of end-to-end (E2E) automatic speech recognition (ASR) models continues to improve as they are scaled to larger sizes, with some now reaching billions of parameters. Widespread deployment and adoption of these models, however, requires computationally efficient strategies for decoding. In the present work, we study one such strategy: applying multiple frame reduction layers in the encoder to compress encoder outputs into a small number of output frames. While similar techniques have been investigated in previous work, we achieve dramatically more reduction than has previously been demonstrated through the use of multiple funnel reduction layers. Through ablations, we study the impact of various architectural choices in the encoder to identify the most effective strategies. We demonstrate that we can generate one encoder output frame for every 2.56 sec of input speech, without significantly affecting word error rate on a large-scale voice search task, while improving encoder and decoder latencies by 48% and 92% respectively, relative to a strong but computationally expensive baseline.