4 citations · 6 across the 9 of their papers we have counts for
10 papers
Audio LLMs Know When They Can't Hear You
Amirhosein Javadi, Richa Dixit, Mehrdad Farajtabar +3
Audio large language models allow users to interact with the model through speech. When an input recording is too degraded, the model may misinterpret the user's query and respond…
Efficient Mixture-of-Experts with Speculative Decoding via Expert Coactivation
Kumari Nishu, Han-Byul Kim, Santosh Chilkunda +13
Mixture-of-Experts (MoE) models are increasingly deployed alongside Speculative Decoding (SD) to accelerate inference, but combining the two is challenging. SD improves the inferen…
Compressing Streaming Neural Audio Encoders via Latent-Space Distillation
Prasanth Yadla, Mohammad Samragh, Dongseong Hwang +7
System-wide Dictation on Apple devices runs entirely on-device, and the speech it transcribes reaches the foundation model through a tokenizer: an encoder that maps short windows o…
SpecMD: A Comprehensive Study On Speculative Expert Prefetching
Duc Hoang, Ajay Jaiswal, Mohammad Samragh +1
Mixture-of-Experts (MoE) models enable sparse expert activation, meaning that only a subset of the model's parameters is used during each inference. However, to translate this spar…
MoEs Are Stronger than You Think: Hyper-Parallel Inference Scaling with RoE
Soheil Zibakhsh, Mohammad Samragh, Kumari Nishu +3
The generation quality of large language models (LLMs) is often improved by utilizing inference-time sequence-level scaling methods (e.g., Chain-of-Thought). We introduce hyper-par…
Your LLM Knows the Future: Uncovering Its Multi-Token Prediction Potential
Mohammad Samragh, Arnav Kundu, David Harrison +4
Autoregressive language models are constrained by their inherently sequential nature, generating one token at a time. This paradigm limits inference speed and parallelism, especial…
SPD: Sync-Point Drop for Efficient Tensor Parallelism of Large Language Models
Han-Byul Kim, Duc Hoang, Arnav Kundu +2
With the rapid expansion in the scale of large language models (LLMs), enabling efficient distributed inference across multiple computing units has become increasingly critical. Ho…
Scaling Smart: Accelerating Large Language Model Pre-training with Small Model Initialization
Mohammad Samragh, Iman Mirzadeh, Keivan Alizadeh Vahid +5
The pre-training phase of language models often begins with randomly initialized parameters. With the current trends in scaling models, training their large number of parameters ca…
Weight subcloning: direct initialization of transformers using larger pretrained ones
Mohammad Samragh, Mehrdad Farajtabar, Sachin Mehta +5
Training large transformer models from scratch for a target task requires lots of data and is computationally demanding. The usual practice of transfer learning overcomes this chal…
Improving vision-inspired keyword spotting using dynamic module skipping in streaming conformer encoder
Alexandre Bittar, Paul Dixon, Mohammad Samragh +2
Using a vision-inspired keyword spotting framework, we propose an architecture with input-dependent dynamic depth capable of processing streaming audio. Specifically, we extend a c…