1.3k citations · 13.3k across the 201 of their papers we have counts for
204 papers
Scaling Properties of Continuous Diffusion Spoken Language Models
Jason Ramapuram, Eeshan Gunesh Dhekane, Amitis Shidani +6
Speech-only spoken language models (SLMs) lag behind text and text-speech models in performance, with recent discrete autoregressive (AR) SLMs indicating significant computational…
DiceHuBERT: Distilling HuBERT with a Self-Supervised Learning Objective
Hyung Gun Chi, Zakaria Aldeneh, Tatiana Likhomanenko +5
We introduce DiceHuBERT, a knowledge distillation framework for compressing HuBERT, a widely used self-supervised learning (SSL)-based speech foundation model. Unlike existing dist…
Speaker-IPL: Unsupervised Learning of Speaker Characteristics with i-Vector based Pseudo-Labels
Zakaria Aldeneh, Takuya Higuchi, Jee-weon Jung +6
Iterative self-training, or iterative pseudo-labeling (IPL) -- using an improved model from the current iteration to provide pseudo-labels for the next iteration -- has proven to b…
Exploring Prediction Targets in Masked Pre-Training for Speech Foundation Models
Li-Wei Chen, Takuya Higuchi, He Bai +6
Speech foundation models, such as HuBERT and its variants, are pre-trained on large amounts of unlabeled speech data and then used for a range of downstream tasks. These models use…
Towards Automatic Assessment of Self-Supervised Speech Models using Rank
Zakaria Aldeneh, Vimal Thilak, Takuya Higuchi +2
This study explores using embedding rank as an unsupervised evaluation metric for general-purpose speech encoders trained via self-supervised learning (SSL). Traditionally, assessi…
Theory, Analysis, and Best Practices for Sigmoid Self-Attention
Jason Ramapuram, Federico Danieli, Eeshan Dhekane +8
Attention is a key part of the transformer architecture. It is a sequence-to-sequence mapping that transforms each sequence element into a weighted sum of values. The weights are t…
Generating Gender Alternatives in Machine Translation
Sarthak Garg, Mozhdeh Gheini, Clara Emmanuel +3
Machine translation (MT) systems often translate terms with ambiguous gender (e.g., English term "the nurse") into the gendered form that is most prevalent in the systems' training…
Can you Remove the Downstream Model for Speaker Recognition with Self-Supervised Speech Features?
Zakaria Aldeneh, Takuya Higuchi, Jee-weon Jung +6
Self-supervised features are typically used in place of filter-bank features in speaker verification models. However, these models were originally designed to ingest filter-bank fe…
Enabling Differentially Private Federated Learning for Speech Recognition: Benchmarks, Adaptive Optimizers and Gradient Clipping
Martin Pelikan, Sheikh Shams Azam, Vitaly Feldman +4
While federated learning (FL) and differential privacy (DP) have been extensively studied, their application to automatic speech recognition (ASR) remains largely unexplored due to…
Importance of Smoothness Induced by Optimizers in FL4ASR: Towards Understanding Federated Learning for End-to-End ASR
Sheikh Shams Azam, Tatiana Likhomanenko, Martin Pelikan +1
In this paper, we start by training End-to-End Automatic Speech Recognition (ASR) models using Federated Learning (FL) and examining the fundamental considerations that can be pivo…
How to Scale Your EMA
Dan Busbridge, Jason Ramapuram, Pierre Ablin +4
Preserving training dynamics across batch sizes is an important tool for practical machine learning as it enables the trade-off between batch size and wall-clock time. This trade-o…
VISION Datasets: A Benchmark for Vision-based InduStrial InspectiON
Haoping Bai, Shancong Mou, Tatiana Likhomanenko +6
Despite progress in vision-based inspection algorithms, real-world industrial challenges -- specifically in data availability, quality, and complex production requirements -- often…
Stabilizing Transformer Training by Preventing Attention Entropy Collapse
Shuangfei Zhai, Tatiana Likhomanenko, Etai Littwin +5
Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In…
Position Prediction as an Effective Pretraining Strategy
Shuangfei Zhai, Navdeep Jaitly, Jason Ramapuram +7
Transformers have gained increasing popularity in a wide range of applications, including Natural Language Processing (NLP), Computer Vision and Speech Recognition, because of thei…
Search for the lepton-flavour violating decays
LHCb collaboration, R. Aaij, C. Abellán Beteta +895
A search for the lepton-flavour violating decays is performed using a sample of proton-proton collision data, collected with the LHCb experiment at ce…
Amplitude analysis of the decay
LHCb collaboration, R. Aaij, C. Abellán Beteta +868
The results of an amplitude analysis of the charmless three-body decay , in which -violation effects are taken into account, are reported. The anal…
Precision measurement of the , and baryon lifetimes
LHCb collaboration, R. Aaij, C. Abellán Beteta +847
We report measurements of the lifetimes of the , and charm baryons using proton-proton collision data at center-of-mass energies of 7 and 8\tev, correspondin…
Measurement of psi(2S) production cross-sections in proton-proton collisions at 7 and 13 TeV
LHCb collaboration, R. Aaij, C. Abellán Beteta +842
The cross-sections of meson production in proton-proton collisions at are measured with a data sample collected by the LHCb detector correspondin…
Measurement of CP violation in the decay and search for the decay
LHCb collaboration, R. Aaij, C. Abellán Beteta +868
A measurement of the time-dependent CP-violating asymmetry in decays is presented. Using a sample of proton-proton collision data corresponding to an integrate…
Observation of the decay
LHCb collaboration, R. Aaij, C. Abellán Beteta +903
Using proton-proton collision data, collected with the LHCb detector and corresponding to 1.0, 2.0 and 1.9fb of integrated luminosity at the centre-of-mass energies of 7, 8,…
Observation of a narrow pentaquark state, , and of two-peak structure of the
LHCb collaboration, R. Aaij, C. Abellán Beteta +866
A narrow pentaquark state, , decaying to is discovered with a statistical significance of in a data sample of decays which is an or…
Amplitude analysis of the decays and measurement of the branching fraction of the decay
LHCb Collaboration, R. Aaij, C. Abellán Beteta +844
The and decays are studied using proton-proton collision data corresponding to an integrated luminosity of 3…
Search for the lepton-flavour-violating decays and
LHCb collaboration, R. Aaij, C. Abellán Beteta +863
A search for and decays is performed using data corresponding to an integrated luminosity of 3 fb of proton-proton coll…
A search for decays
LHCb Collaboration, R. Aaij, C. Abellán Beteta +848
A search for the baryon through the decay is performed with a data sample corresponding to an integrated luminosit…