no code implementations • Findings (ACL) 2022 • Zihan Wang, Jiuxiang Gu, Jason Kuen, Handong Zhao, Vlad Morariu, Ruiyi Zhang, Ani Nenkova, Tong Sun, Jingbo Shang
We present a comprehensive study of sparse attention patterns in Transformer models.
no code implementations • EMNLP 2021 • Zihan Wang, chengyu dong, Jingbo Shang
In this paper, we present an empirical property of these representations—”average” approximates “first principal component”.
no code implementations • 21 Dec 2022 • Zihan Wang, Naoki Yoshinaga
Therefore, in this study, we introduce a task of generating game commentaries from structured data records to address the problem.
no code implementations • 7 Dec 2022 • Zihan Wang, Jason Lee, Qi Lei
Understanding when and how much a model gradient leaks information about the training sample is an important question in privacy.
no code implementations • 14 Nov 2022 • Weijie Sun, Sunil Vasu Kalmady, Nariman Sepehrvand, Luan Manh Chu, Zihan Wang, Amir Salimi, Abram Hindle, Russell Greiner, Padma Kaul
Pandemic outbreaks such as COVID-19 occur unexpectedly, and need immediate action due to their potential devastating consequences on global health.
no code implementations • 31 Oct 2022 • Zihan Wang, Qi Meng, HaiFeng Lan, Xinrui Zhang, Kehao Guo, Akshat Gupta
While Speech Emotion Recognition (SER) is a common application for popular languages, it continues to be a problem for low-resourced languages, i. e., languages with no pretrained speech-to-text recognition models.
2 code implementations • 5 Oct 2022 • Aohan Zeng, Xiao Liu, Zhengxiao Du, Zihan Wang, Hanyu Lai, Ming Ding, Zhuoyi Yang, Yifan Xu, Wendi Zheng, Xiao Xia, Weng Lam Tam, Zixuan Ma, Yufei Xue, Jidong Zhai, WenGuang Chen, Peng Zhang, Yuxiao Dong, Jie Tang
We introduce GLM-130B, a bilingual (English and Chinese) pre-trained language model with 130 billion parameters.
Ranked #1 on
Language Modelling
on CLUE (CMRC2018)
no code implementations • 5 Oct 2022 • Yufan Zhuang, Zihan Wang, Fangbo Tao, Jingbo Shang
We propose Waveformer that learns attention mechanism in the wavelet coefficient space, requires only linear time complexity, and enjoys universal approximating power.
no code implementations • 16 Sep 2022 • Yilun Hao, Ruinan Wang, Zhangjie Cao, Zihan Wang, Yuchen Cui, Dorsa Sadigh
Specifically, we design a masked policy network with a binary mask to block certain modalities.
1 code implementation • 15 Sep 2022 • Pingyi Hu, Zihan Wang, Ruoxi Sun, Hu Wang, Minhui Xue
To achieve this, we propose Multi-modal Models Membership Inference (M^4I) with two attack methods to infer the membership status, named metric-based (MB) M^4I and feature-based (FB) M^4I, respectively.
1 code implementation • 25 Jun 2022 • Akide Liu, Zihan Wang
This competition focus on Urban-Sense Segmentation based on the vehicle camera view.
1 code implementation • 24 Jun 2022 • Zihan Wang, Na Huang, Fei Sun, Pengjie Ren, Zhumin Chen, Hengliang Luo, Maarten de Rijke, Zhaochun Ren
To address the above limitations, we propose a Debiasing Learning for Membership Inference Attacks against recommender systems (DL-MIA) framework that has four main components: (1) a difference vector generator, (2) a disentangled encoder, (3) a weight estimator, and (4) an attack model.
no code implementations • 4 Jun 2022 • Zihan Wang, Ruimin Chen, Mengxuan Liu, Guanfang Dong, Anup Basu
We propose a method SPGNet for 3D human pose estimation that mixes multi-dimensional re-projection into supervised learning.
Ranked #37 on
3D Human Pose Estimation
on Human3.6M
1 code implementation • 28 May 2022 • Ziang Li, Ming Ding, Weikai Li, Zihan Wang, Ziyu Zeng, Yukuo Cen, Jie Tang
graph benchmark (IGB) consisting of 4 datasets.
1 code implementation • 24 May 2022 • Zihan Wang, Kewen Zhao, Zilong Wang, Jingbo Shang
Fine-tuning pre-trained language models has recently become a common practice in building NLP models for various tasks, especially few-shot tasks.
no code implementations • 24 May 2022 • Lesheng Jin, Zihan Wang, Jingbo Shang
Inspired by this observation, in WeDef, we define the reliability of samples based on whether the predictions of the weak classifier agree with their labels in the poisoned training set.
1 code implementation • ACL 2022 • Jinyu Guo, Kai Shuang, Jijie Li, Zihan Wang, Yixuan Liu
However, no matter how the dialogue history is used, each existing model uses its own consistent dialogue history during the entire state tracking process, regardless of which slot is updated.
no code implementations • 9 May 2022 • Zihan Wang, Gang Wu, Yan Wang
The RNN often used in previous work is not suitable to process short sessions, because RNN only focuses on the sequential relationship, which we find is not the only relationship between items in short sessions.
1 code implementation • 28 Apr 2022 • Zihan Wang, Peiyi Wang, Tianyu Liu, Binghuai Lin, Yunbo Cao, Zhifang Sui, Houfeng Wang
However, in this paradigm, there exists a huge gap between the classification tasks with sophisticated label hierarchy and the masked language model (MLM) pretraining tasks of PLMs and thus the potentials of PLMs can not be fully tapped.
1 code implementation • ACL 2022 • Zihan Wang, Peiyi Wang, Lianzhe Huang, Xin Sun, Houfeng Wang
Hierarchical text classification is a challenging subtask of multi-label classification due to its complex label hierarchy.
no code implementations • 2 Mar 2022 • Zihan Wang, Zhangjie Cao, Yilun Hao, Dorsa Sadigh
Correspondence learning is a fundamental problem in robotics, which aims to learn a mapping between state, action pairs of agents of different dynamics or embodiments.
no code implementations • 7 Feb 2022 • Zhangjie Cao, Zihan Wang, Dorsa Sadigh
Existing learning from demonstration algorithms usually assume access to expert demonstrations.
1 code implementation • 9 Nov 2021 • Zihan Wang, Jialin Lu, Oliver Snow, Martin Ester
Despite recent progress in artificial intelligence and machine learning, many state-of-the-art methods suffer from a lack of explainability and transparency.
1 code implementation • 16 Sep 2021 • Minxing Zhang, Zhaochun Ren, Zihan Wang, Pengjie Ren, Zhumin Chen, Pengfei Hu, Yang Zhang
In this paper, we make the first attempt on quantifying the privacy leakage of recommender systems through the lens of membership inference.
no code implementations • ACL 2021 • Jinyu Guo, Kai Shuang, Jijie Li, Zihan Wang
However, the overwhelming majority of the slots in each turn should simply inherit the slot values from the previous turn.
2 code implementations • 28 May 2021 • Xiaotao Gu, Zihan Wang, Zhenyu Bi, Yu Meng, Liyuan Liu, Jiawei Han, Jingbo Shang
Training a conventional neural tagger based on silver labels usually faces the risk of overfitting phrase surface names.
Ranked #1 on
Phrase Tagging
on KPTimes
no code implementations • 13 May 2021 • Zihan Wang, Hongye Song, Zhaochun Ren, Pengjie Ren, Zhumin Chen, Xiaozhong Liu, Hongsong Li, Maarten de Rijke
First, contract elements are far more fine-grained than named entities, which hinders the transfer of extractors.
Cross-Domain Named Entity Recognition
named-entity-recognition
+1
1 code implementation • 22 Apr 2021 • Runlong Yu, Yuyang Ye, Qi Liu, Zihan Wang, Chunfeng Yang, Yucheng Hu, Enhong Chen
Motivated by this, we propose a novel Extreme Cross Network, abbreviated XCrossNet, which aims at learning dense and sparse feature interactions in an explicit manner.
1 code implementation • 18 Apr 2021 • Zihan Wang, chengyu dong, Jingbo Shang
In this paper, we present an empirical property of these representations -- "average" approximates "first principal component".
no code implementations • 6 Nov 2020 • Zhendong Ai, Zihan Wang, Wei Cui
The ECG monitoring device, abbreviated as ECGM, is designed based on ferroelectric microprocessor which provides ultra-low power consumption and contains four parts-MCU, BLE, Sensors and Power.
3 code implementations • NAACL 2021 • Zihan Wang, Dheeraj Mekala, Jingbo Shang
Finally, we pick the most confident documents from each cluster to train a text classifier.
no code implementations • 10 Sep 2020 • Sarah E. Finch, James D. Finch, Ali Ahmadvand, Ingyu, Choi, Xiangjue Dong, Ruixiang Qi, Harshita Sahijwani, Sergey Volokhin, Zihan Wang, ZiHao Wang, Jinho D. Choi
Inspired by studies on the overwhelming presence of experience-sharing in human-human conversations, Emora, the social chatbot developed by Emory University, aims to bring such experience-focused interaction to the current field of conversational AI.
no code implementations • Findings of the Association for Computational Linguistics 2020 • Zihan Wang, Karthikeyan K, Stephen Mayhew, Dan Roth
Multilingual BERT (M-BERT) has been a huge success in both supervised and zero-shot cross-lingual transfer learning.
no code implementations • ICLR 2020 • Karthikeyan K, Zihan Wang, Stephen Mayhew, Dan Roth
Recent work has exhibited the surprising cross-lingual abilities of multilingual BERT (M-BERT) -- surprising since it is trained without any cross-lingual objective and with no aligned data.
no code implementations • 11 Nov 2019 • Yunan Zhang, Xiang Cheng, Yufeng Zhang, Zihan Wang, Zhengqi Fang, Xiaoyan Wang, Zhenya Huang, ChengXiang Zhai
Answering complex questions involving multiple entities and relations is a challenging task.
1 code implementation • IJCNLP 2019 • Zihan Wang, Jingbo Shang, Liyuan Liu, Lihao Lu, Jiacheng Liu, Jiawei Han
Therefore, we manually correct these label mistakes and form a cleaner test set.
Ranked #3 on
Named Entity Recognition
on CoNLL++
(using extra training data)
1 code implementation • 20 Aug 2019 • Yu Meng, Jiaxin Huang, Guangyuan Wang, Zihan Wang, Chao Zhang, Yu Zhang, Jiawei Han
We propose a new task, discriminative topic mining, which leverages a set of user-provided category names to mine discriminative topics from text corpora.
1 code implementation • 14 Aug 2019 • Liyuan Liu, Zihan Wang, Jingbo Shang, Dandong Yin, Heng Ji, Xiang Ren, Shaowen Wang, Jiawei Han
Our model neither requires the conversion from character sequences to word sequences, nor assumes tokenizer can correctly detect all word boundaries.
no code implementations • 11 May 2019 • Zihan Wang, Yaoguang Li, Wei Cui
By applying various existing learning methods to our ECG dataset, we find that current methods which can well support the identification of individuals under rests, do not suffice to present satisfying ECGID performance under exercise situations, therefore exposing the deficiency of existing ECG identification methods.
no code implementations • 11 Oct 2018 • Homanga Bharadhwaj, Zihan Wang, Yoshua Bengio, Liam Paull
Learning effective visuomotor policies for robots purely from data is challenging, but also appealing since a learning-based system should not require manual tuning or calibration.