Picture for Maha Elbayad

Maha Elbayad

NLLB Team

Merging Text Transformer Models from Different Initializations

Add code
Mar 07, 2024
Figure 1 for Merging Text Transformer Models from Different Initializations
Figure 2 for Merging Text Transformer Models from Different Initializations
Figure 3 for Merging Text Transformer Models from Different Initializations
Figure 4 for Merging Text Transformer Models from Different Initializations
Viaarxiv icon

SpiRit-LM: Interleaved Spoken and Written Language Model

Add code
Feb 08, 2024
Viaarxiv icon

Seamless: Multilingual Expressive and Streaming Speech Translation

Add code
Dec 08, 2023
Figure 1 for Seamless: Multilingual Expressive and Streaming Speech Translation
Figure 2 for Seamless: Multilingual Expressive and Streaming Speech Translation
Figure 3 for Seamless: Multilingual Expressive and Streaming Speech Translation
Figure 4 for Seamless: Multilingual Expressive and Streaming Speech Translation
Viaarxiv icon

Added Toxicity Mitigation at Inference Time for Multimodal and Massively Multilingual Translation

Add code
Nov 11, 2023
Figure 1 for Added Toxicity Mitigation at Inference Time for Multimodal and Massively Multilingual Translation
Figure 2 for Added Toxicity Mitigation at Inference Time for Multimodal and Massively Multilingual Translation
Figure 3 for Added Toxicity Mitigation at Inference Time for Multimodal and Massively Multilingual Translation
Figure 4 for Added Toxicity Mitigation at Inference Time for Multimodal and Massively Multilingual Translation
Viaarxiv icon

SeamlessM4T-Massively Multilingual & Multimodal Machine Translation

Add code
Aug 23, 2023
Figure 1 for SeamlessM4T-Massively Multilingual & Multimodal Machine Translation
Figure 2 for SeamlessM4T-Massively Multilingual & Multimodal Machine Translation
Figure 3 for SeamlessM4T-Massively Multilingual & Multimodal Machine Translation
Figure 4 for SeamlessM4T-Massively Multilingual & Multimodal Machine Translation
Viaarxiv icon

Towards Being Parameter-Efficient: A Stratified Sparsely Activated Transformer with Dynamic Capacity

Add code
May 03, 2023
Viaarxiv icon

Efficiently Upgrading Multilingual Machine Translation Models to Support More Languages

Add code
Feb 07, 2023
Figure 1 for Efficiently Upgrading Multilingual Machine Translation Models to Support More Languages
Figure 2 for Efficiently Upgrading Multilingual Machine Translation Models to Support More Languages
Figure 3 for Efficiently Upgrading Multilingual Machine Translation Models to Support More Languages
Figure 4 for Efficiently Upgrading Multilingual Machine Translation Models to Support More Languages
Viaarxiv icon

Fixing MoE Over-Fitting on Low-Resource Languages in Multilingual Machine Translation

Add code
Dec 15, 2022
Figure 1 for Fixing MoE Over-Fitting on Low-Resource Languages in Multilingual Machine Translation
Figure 2 for Fixing MoE Over-Fitting on Low-Resource Languages in Multilingual Machine Translation
Figure 3 for Fixing MoE Over-Fitting on Low-Resource Languages in Multilingual Machine Translation
Figure 4 for Fixing MoE Over-Fitting on Low-Resource Languages in Multilingual Machine Translation
Viaarxiv icon

Causes and Cures for Interference in Multilingual Translation

Add code
Dec 14, 2022
Viaarxiv icon

No Language Left Behind: Scaling Human-Centered Machine Translation

Add code
Jul 11, 2022
Figure 1 for No Language Left Behind: Scaling Human-Centered Machine Translation
Figure 2 for No Language Left Behind: Scaling Human-Centered Machine Translation
Figure 3 for No Language Left Behind: Scaling Human-Centered Machine Translation
Figure 4 for No Language Left Behind: Scaling Human-Centered Machine Translation
Viaarxiv icon