arXiv cs.CL· Nir Krakowski, Tamir Cohen, Ohad Zohar, Chen Zarfati·· 12 小时前AI 评分40
E-Gaborformer：采用 Gabor 字典卷积与 Delta 规则记忆的严格因果语音识别器
Gabor Gabor: E-Gaborformer, a Strictly Causal Speech Recognizer with Gabor-Dictionary Convolutions and Delta-Rule Memory
AI 导读
E-Gaborformer 是一种严格因果的语音识别编码器，将 Gated DeltaNet-2 循环层与滑动窗口注意力按 3:1 比例组合，所有时间卷积均由固定的因果 Gabor 原子字典构建，流式状态恒定，生产规模下每次调用约 19 MB（含预测网络），而同等规模的因果注意力编码器十分钟后累积超过 1 GB。
正文 · 原文
Abstract:Real-time voice agents need a speech recognizer that never waits for future audio and whose memory per call does not grow with the length of the call. We present E-Gaborformer, a strictly causal encoder that combines Gated DeltaNet-2 recurrent layers with sliding-window attention in a 3:1 ratio, and builds every temporal convolution from a fixed dictionary of causal Gabor atoms. Its streaming state is constant, about 19 MB per call at production scale including the prediction network, where a causal-attention encoder of the same size accumulates over 1 GB after ten minutes. In a controlled causal study on LibriSpeech, E-Gaborformer leaves the initial plateau earlier, reaches a given accuracy up to 18% sooner, and matches or beats an E-Branchformer trained with the identical front end and recipe. At production scale, a 0.6B-parameter Hebrew and English streaming transducer built on E-Gaborformer has 28 to 31% lower word error rate (WER) than NVIDIA's Nemotron 3.5 ASR Streaming, a model of the same size, on English voice-agent traffic, and up to 20% lower Hebrew WER than our previous production system. The same weights serve strictly causal and 480 ms-lookahead decoding with no loss in accuracy, and a two-class formulation of the RNN-T lattice trains the transducer 2.6x faster with exact loss and gradients.
| Comments: | 19 pages, 6 figures, 9 tables |
| Subjects: | Machine Learning (cs.LG); Computation and Language (cs.CL); Sound (cs.SD) |
| ACM classes: | I.2.7; I.5.4 |
| Cite as: | arXiv:2611.08092 [cs.LG] |
| (or arXiv:2611.08092v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08092 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Nir Krakowski [view email]
[v1]
Thu, 5 Nov 2026 14:36:34 UTC (32 KB)
来源：arXiv cs.CL · arxiv.org