­

【清華大學NLP】預訓練語言模型（PLM）必讀論文清單，附論文PDF、源碼和模型鏈接

  • 2019 年 10 月 11 日
  • 筆記

【導讀】近兩年來，ELMO、BERT等預訓練語言模型（PLM）在多項任務中刷新了榜單，引起了學術界和工業界的大量關注。本文介紹清華大學NLP給出的預訓練語言模型必讀論文清單，包含論文的PDF鏈接、源碼和模型等。

清華大學NLP在Github項目thunlp/PLMpapers中提供了預訓練語言模型必讀論文清單，包含了論文的PDF鏈接、源碼和模型等，具體清單如下：

模型：

  1. Deep contextualized word representations. Matthew E. Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee and Luke Zettlemoyer. NAACL 2018.
    • 論文: https://arxiv.org/pdf/1802.05365.pdf
    • 工程: https://allennlp.org/elmo (ELMo)
  2. Universal Language Model Fine-tuning for Text Classification. Jeremy Howard and Sebastian Ruder. ACL 2018.
    • 論文: https://www.aclweb.org/anthology/P18-1031
    • 工程: http://nlp.fast.ai/category/classification.html (ULMFiT)
  3. Improving Language Understanding by Generative Pre-Training. Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever. Preprint.
    • 論文: https://s3-us-west-2.amazonaws.com/openai-assets/research-covers/language-unsupervised/language_understanding_paper.pdf
    • 工程: https://openai.com/blog/language-unsupervised/ (GPT)
  4. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. Jacob Devlin, Ming-Wei Chang, Kenton Lee and Kristina Toutanova. NAACL 2019.
    • 論文: https://arxiv.org/pdf/1810.04805.pdf
    • 代碼+模型: https://github.com/google-research/bert
  5. Language Models are Unsupervised Multitask Learners. Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei and Ilya Sutskever. Preprint.
    • 論文: https://d4mucfpksywv.cloudfront.net/better-language-models/language_models_are_unsupervised_multitask_learners.pdf
    • 代碼: https://github.com/openai/gpt-2 (GPT-2)
  6. ERNIE: Enhanced Language Representation with Informative Entities. Zhengyan Zhang, Xu Han, Zhiyuan Liu, Xin Jiang, Maosong Sun and Qun Liu. ACL2019.
    • 論文: https://www.aclweb.org/anthology/P19-1139
    • 代碼+模型: https://github.com/thunlp/ERNIE (ERNIE (Tsinghua) )
  7. ERNIE: Enhanced Representation through Knowledge Integration. Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian and Hua Wu. Preprint.
    • 論文: https://arxiv.org/pdf/1904.09223.pdf
    • 代碼: https://github.com/PaddlePaddle/ERNIE/tree/develop/ERNIE (ERNIE (Baidu) )
  8. Defending Against Neural Fake News. Rowan Zellers, Ari Holtzman, Hannah Rashkin, Yonatan Bisk, Ali Farhadi, Franziska Roesner, Yejin Choi. NeurIPS.
    • 論文: https://arxiv.org/pdf/1905.12616.pdf
    • 工程: https://rowanzellers.com/grover/ (Grover)
  9. Cross-lingual Language Model Pretraining. Guillaume Lample, Alexis Conneau. NeurIPS2019.
    • 論文: https://arxiv.org/pdf/1901.07291.pdf
    • 代碼+模型: https://github.com/facebookresearch/XLM (XLM)
  10. Multi-Task Deep Neural Networks for Natural Language Understanding. Xiaodong Liu, Pengcheng He, Weizhu Chen, Jianfeng Gao. ACL2019.
    • 論文: https://www.aclweb.org/anthology/P19-1441
    • 代碼+模型: https://github.com/namisan/mt-dnn (MT-DNN)
  11. MASS: Masked Sequence to Sequence Pre-training for Language Generation. Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu. ICML2019.
    • 論文: https://arxiv.org/pdf/1905.02450.pdf
    • 代碼+模型: https://github.com/microsoft/MASS
  12. Unified Language Model Pre-training for Natural Language Understanding and Generation. Li Dong, Nan Yang, Wenhui Wang, Furu Wei, Xiaodong Liu, Yu Wang, Jianfeng Gao, Ming Zhou, Hsiao-Wuen Hon. Preprint.
    • 論文: https://arxiv.org/pdf/1905.03197.pdf (UniLM)
  13. XLNet: Generalized Autoregressive Pretraining for Language Understanding. Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Ruslan Salakhutdinov, Quoc V. Le. NeurIPS2019.
    • 論文: https://arxiv.org/pdf/1906.08237.pdf
    • 代碼+模型: https://github.com/zihangdai/xlnet
  14. RoBERTa: A Robustly Optimized BERT Pretraining Approach. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov. Preprint.
    • 論文: https://arxiv.org/pdf/1907.11692.pdf
    • 代碼+模型: https://github.com/pytorch/fairseq
  15. SpanBERT: Improving Pre-training by Representing and Predicting Spans. Mandar Joshi, Danqi Chen, Yinhan Liu, Daniel S. Weld, Luke Zettlemoyer, Omer Levy. Preprint.
    • 論文: https://arxiv.org/pdf/1907.10529.pdf
    • 代碼+模型: https://github.com/facebookresearch/SpanBERT
  16. Knowledge Enhanced Contextual Word Representations. Matthew E. Peters, Mark Neumann, Robert L. Logan IV, Roy Schwartz, Vidur Joshi, Sameer Singh, Noah A. Smith. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1909.04164.pdf (KnowBert)
  17. VisualBERT: A Simple and Performant Baseline for Vision and Language. Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang. Preprint.
    • 論文: https://arxiv.org/pdf/1908.03557.pdf
    • 代碼+模型: https://github.com/uclanlp/visualbert
  18. ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks. Jiasen Lu, Dhruv Batra, Devi Parikh, Stefan Lee. NeurIPS.
    • 論文: https://arxiv.org/pdf/1908.02265.pdf
    • 代碼+模型: https://github.com/jiasenlu/vilbert_beta
  19. VideoBERT: A Joint Model for Video and Language Representation Learning. Chen Sun, Austin Myers, Carl Vondrick, Kevin Murphy, Cordelia Schmid. ICCV2019.
    • 論文: https://arxiv.org/pdf/1904.01766.pdf
  20. LXMERT: Learning Cross-Modality Encoder Representations from Transformers. Hao Tan, Mohit Bansal. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1908.07490.pdf
    • 代碼+模型: https://github.com/airsplay/lxmert
  21. VL-BERT: Pre-training of Generic Visual-Linguistic Representations. Weijie Su, Xizhou Zhu, Yue Cao, Bin Li, Lewei Lu, Furu Wei, Jifeng Dai. Preprint.
    • 論文: https://arxiv.org/pdf/1908.08530.pdf
  22. Unicoder-VL: A Universal Encoder for Vision and Language by Cross-modal Pre-training. Gen Li, Nan Duan, Yuejian Fang, Ming Gong, Daxin Jiang, Ming Zhou. Preprint.
    • 論文: https://arxiv.org/pdf/1908.06066.pdf
  23. K-BERT: Enabling Language Representation with Knowledge Graph. Weijie Liu, Peng Zhou, Zhe Zhao, Zhiruo Wang, Qi Ju, Haotang Deng, Ping Wang. Preprint.
    • 論文: https://arxiv.org/pdf/1909.07606.pdf
  24. Fusion of Detected Objects in Text for Visual Question Answering. Chris Alberti, Jeffrey Ling, Michael Collins, David Reitter. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1908.05054.pdf (B2T2)
  25. Contrastive Bidirectional Transformer for Temporal Representation Learning. Chen Sun, Fabien Baradel, Kevin Murphy, Cordelia Schmid. Preprint.
    • 論文: https://arxiv.org/pdf/1906.05743.pdf (CBT)
  26. ERNIE 2.0: A Continual Pre-training Framework for Language Understanding. Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Hao Tian, Hua Wu, Haifeng Wang. Preprint.
    • 論文: https://arxiv.org/pdf/1907.12412v1.pdf
    • 代碼: https://github.com/PaddlePaddle/ERNIE/blob/develop/README.md
  27. 75 Languages, 1 Model: Parsing Universal Dependencies Universally. Dan Kondratyuk, Milan Straka. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1904.02099.pdf
    • 代碼+模型: https://github.com/hyperparticle/udify (UDify)
  28. Pre-Training with Whole Word Masking for Chinese BERT. Yiming Cui, Wanxiang Che, Ting Liu, Bing Qin, Ziqing Yang, Shijin Wang, Guoping Hu. Preprint.
    • 論文: https://arxiv.org/pdf/1906.08101.pdf
    • 代碼+模型: https://github.com/ymcui/Chinese-BERT-wwm/blob/master/README_EN.md (Chinese-BERT-wwm)

知識蒸餾和模型壓縮:

  1. TinyBERT: Distilling BERT for Natural Language Understanding. Xiaoqi Jiao, Yichun Yin, Lifeng Shang, Xin Jiang, Xiao Chen, Linlin Li, Fang Wang, Qun Liu.
    • 論文: https://arxiv.org/pdf/1909.10351v1.pdf
  2. Distilling Task-Specific Knowledge from BERT into Simple Neural Networks. Raphael Tang, Yao Lu, Linqing Liu, Lili Mou, Olga Vechtomova, Jimmy Lin. Preprint.
    • 論文: https://arxiv.org/pdf/1903.12136.pdf
  3. Patient Knowledge Distillation for BERT Model Compression. Siqi Sun, Yu Cheng, Zhe Gan, Jingjing Liu. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1908.09355.pdf
    • 代碼: https://github.com/intersun/PKD-for-BERT-Model-Compression
  4. Model Compression with Multi-Task Knowledge Distillation for Web-scale Question Answering System. Ze Yang, Linjun Shou, Ming Gong, Wutao Lin, Daxin Jiang. Preprint.
    • 論文: https://arxiv.org/pdf/1904.09636.pdf
  5. PANLP at MEDIQA 2019: Pre-trained Language Models, Transfer Learning and Knowledge Distillation. Wei Zhu, Xiaofeng Zhou, Keqiang Wang, Xun Luo, Xiepeng Li, Yuan Ni, Guotong Xie. The 18th BioNLP workshop.
    • 論文: https://www.aclweb.org/anthology/W19-5040
  6. Improving Multi-Task Deep Neural Networks via Knowledge Distillation for Natural Language Understanding. Xiaodong Liu, Pengcheng He, Weizhu Chen, Jianfeng Gao. Preprint.
    • 論文: https://arxiv.org/pdf/1904.09482.pdf
    • 代碼+模型: https://github.com/namisan/mt-dnn
  7. Well-Read Students Learn Better: The Impact of Student Initialization on Knowledge Distillation. Iulia Turc, Ming-Wei Chang, Kenton Lee, Kristina Toutanova. Preprint.
    • 論文: https://arxiv.org/pdf/1908.08962.pdf
  8. Small and Practical BERT Models for Sequence Labeling. Henry Tsai, Jason Riesa, Melvin Johnson, Naveen Arivazhagan, Xin Li, Amelia Archer. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1909.00100.pdf
  9. Q-BERT: Hessian Based Ultra Low Precision Quantization of BERT. Sheng Shen, Zhen Dong, Jiayu Ye, Linjian Ma, Zhewei Yao, Amir Gholami, Michael W. Mahoney, Kurt Keutzer. Preprint.
    • 論文: https://arxiv.org/pdf/1909.05840.pdf
  10. ALBERT: A Lite BERT for Self-supervised Learning of Language Representations. Anonymous authors. ICLR2020 under review.
    • 論文: https://openreview.net/pdf?id=H1eA7AEtvS

分析:

  1. Revealing the Dark Secrets of BERT. Olga Kovaleva, Alexey Romanov, Anna Rogers, Anna Rumshisky. EMNLP2019.
    • 論文: https://arxiv.org/abs/1908.08593
  2. How Does BERT Answer Questions? A Layer-Wise Analysis of Transformer Representations. Betty van Aken, Benjamin Winter, Alexander Löser, Felix A. Gers. CIKM2019.
    • 論文: https://arxiv.org/pdf/1909.04925.pdf
  3. Are Sixteen Heads Really Better than One?. Paul Michel, Omer Levy, Graham Neubig. Preprint.
    • 論文: https://arxiv.org/pdf/1905.10650.pdf
    • 代碼: https://github.com/pmichel31415/are-16-heads-really-better-than-1
  4. Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailment. Di Jin, Zhijing Jin, Joey Tianyi Zhou, Peter Szolovits. Preprint.
    • 論文: https://arxiv.org/pdf/1907.11932.pdf
    • 代碼: https://github.com/jind11/TextFooler
  5. BERT has a Mouth, and It Must Speak: BERT as a Markov Random Field Language Model. Alex Wang, Kyunghyun Cho. NeuralGen2019.
    • 論文: https://arxiv.org/pdf/1902.04094.pdf
    • 代碼: https://github.com/nyu-dl/bert-gen
  6. Linguistic Knowledge and Transferability of Contextual Representations. Nelson F. Liu, Matt Gardner, Yonatan Belinkov, Matthew E. Peters, Noah A. Smith. NAACL2019.
    • 論文: https://www.aclweb.org/anthology/N19-1112
  7. What Does BERT Look At? An Analysis of BERT's Attention. Kevin Clark, Urvashi Khandelwal, Omer Levy, Christopher D. Manning. BlackBoxNLP2019.
    • 論文: https://arxiv.org/pdf/1906.04341.pdf
    • 代碼: https://github.com/clarkkev/attention-analysis
  8. Open Sesame: Getting Inside BERT's Linguistic Knowledge. Yongjie Lin, Yi Chern Tan, Robert Frank. BlackBoxNLP2019.
    • 論文: https://arxiv.org/pdf/1906.01698.pdf
    • 代碼: https://github.com/yongjie-lin/bert-opensesame
  9. Analyzing the Structure of Attention in a Transformer Language Model. Jesse Vig, Yonatan Belinkov. BlackBoxNLP2019.
    • 論文: https://arxiv.org/pdf/1906.04284.pdf
  10. Blackbox meets blackbox: Representational Similarity and Stability Analysis of Neural Language Models and Brains. Samira Abnar, Lisa Beinborn, Rochelle Choenni, Willem Zuidema. BlackBoxNLP2019.
    • 論文: https://arxiv.org/pdf/1906.01539.pdf
  11. BERT Rediscovers the Classical NLP Pipeline. Ian Tenney, Dipanjan Das, Ellie Pavlick. ACL2019.
    • 論文: https://www.aclweb.org/anthology/P19-1452
  12. How multilingual is Multilingual BERT?. Telmo Pires, Eva Schlinger, Dan Garrette. ACL2019.
    • 論文: https://www.aclweb.org/anthology/P19-1493
  13. What Does BERT Learn about the Structure of Language?. Ganesh Jawahar, Benoît Sagot, Djamé Seddah. ACL2019.
    • 論文: https://www.aclweb.org/anthology/P19-1356
  14. Beto, Bentz, Becas: The Surprising Cross-Lingual Effectiveness of BERT. Shijie Wu, Mark Dredze. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1904.09077.pdf
  15. How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings. Kawin Ethayarajh. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1909.00512.pdf
  16. Probing Neural Network Comprehension of Natural Language Arguments. Timothy Niven, Hung-Yu Kao. ACL2019.
    • 論文: https://www.aclweb.org/anthology/P19-1459
    • 代碼: https://github.com/IKMLab/arct2
  17. Universal Adversarial Triggers for Attacking and Analyzing NLP. Eric Wallace, Shi Feng, Nikhil Kandpal, Matt Gardner, Sameer Singh. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1908.07125.pdf
    • 代碼: https://github.com/Eric-Wallace/universal-triggers
  18. The Bottom-up Evolution of Representations in the Transformer: A Study with Machine Translation and Language Modeling Objectives. Elena Voita, Rico Sennrich, Ivan Titov. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1909.01380.pdf
  19. Do NLP Models Know Numbers? Probing Numeracy in Embeddings. Eric Wallace, Yizhong Wang, Sujian Li, Sameer Singh, Matt Gardner. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1909.07940.pdf
  20. Investigating BERT's Knowledge of Language: Five Analysis Methods with NPIs. Alex Warstadt, Yu Cao, Ioana Grosu, Wei Peng, Hagen Blix, Yining Nie, Anna Alsop, Shikha Bordia, Haokun Liu, Alicia Parrish, Sheng-Fu Wang, Jason Phang, Anhad Mohananey, Phu Mon Htut, Paloma Jeretič, Samuel R. Bowman. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1909.02597.pdf
    • 代碼: https://github.com/alexwarstadt/data_generation
  21. Visualizing and Understanding the Effectiveness of BERT. Yaru Hao, Li Dong, Furu Wei, Ke Xu. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1908.05620.pdf
  22. Visualizing and Measuring the Geometry of BERT. Andy Coenen, Emily Reif, Ann Yuan, Been Kim, Adam Pearce, Fernanda Viégas, Martin Wattenberg. NeurIPS2019.
    • 論文: https://arxiv.org/pdf/1906.02715.pdf
  23. On the Validity of Self-Attention as Explanation in Transformer Models. Gino Brunner, Yang Liu, Damián Pascual, Oliver Richter, Roger Wattenhofer. Preprint.
    • 論文: https://arxiv.org/pdf/1908.04211.pdf
  24. Transformer Dissection: An Unified Understanding for Transformer's Attention via the Lens of Kernel. Yao-Hung Hubert Tsai, Shaojie Bai, Makoto Yamada, Louis-Philippe Morency, Ruslan Salakhutdinov. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1908.11775.pdf
  25. Language Models as Knowledge Bases? Fabio Petroni, Tim Rocktäschel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, Alexander H. Miller, Sebastian Riedel. EMNLP2019.
    • 論文: https://arxiv.org/pdf/1909.01066.pdf
    • 代碼: https://github.com/facebookresearch/LAMA

參考鏈接：

  • https://github.com/thunlp/PLMpapers

-END-

VirMach 便宜 VPS

QNews

QNews