activity
20172022
most citedMulti-Task Deep Neural Networks for Natural Language Understanding

221 citations · 703 across the 22 of their papers we have counts for

collaborators

29 papers

cs.CL20221 cited

AdaMix: Mixture-of-Adaptations for Parameter-efficient Model Tuning

Yaqing Wang, Sahaj Agarwal, Subhabrata Mukherjee +4

Standard fine-tuning of large pre-trained language models (PLMs) for downstream tasks requires updating hundreds of millions to billions of parameters, and storing a large copy of…

cs.CL20221 cited

Open-domain Question Answering via Chain of Reasoning over Heterogeneous Knowledge

Kaixin Ma, Hao Cheng, Xiaodong Liu +2

We propose a novel open-domain question answering (ODQA) framework for answering single/multi-hop questions across heterogeneous knowledge sources. The key novelty of our method is…

cs.CV20228 cited

PTSEFormer: Progressive Temporal-Spatial Enhanced TransFormer Towards Video Object Detection

Han Wang, Jun Tang, Xiaodong Liu +3

Recent years have witnessed a trend of applying context frames to boost the performance of object detection as video object detection. Existing methods usually aggregate features a…

cs.LG202219 cited

METRO: Efficient Denoising Pretraining of Large Scale Autoencoding Language Models with Model Generated Signals

Payal Bajaj, Chenyan Xiong, Guolin Ke +7

We present an efficient method of pretraining large-scale autoencoding language models using training signals generated by an auxiliary model. Originated in ELECTRA, this training…

cs.LG202222 cited

Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer

Greg Yang, Edward J. Hu, Igor Babuschkin +7

Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovere…

cs.CL20223 cited

AutoDistil: Few-shot Task-agnostic Neural Architecture Search for Distilling Large Language Models

Dongkuan Xu, Subhabrata Mukherjee, Xiaodong Liu +5

Knowledge distillation (KD) methods compress large models into smaller students with manually-designed student architectures given pre-specified computational cost. This requires s…