"We have taken note of the questioning of the article "A Roadmap for Big Model" and are verifying the relevant situation, and the Zhiyuan Research Institute encourages academic innovation and academic exchanges, and has zero tolerance for academic misconduct, and the relevant progress will be reported as soon as possible." On April 13, beijing Zhiyuan Artificial Intelligence Research Institute replied to a famous google brain scientist Nicholas Carlini's allegations of alleged plagiarism of a large-scale Academic paper in Sino-foreign cooperation.
Earlier, on April 8, Nicholas Carlini publicly wrote alleging that an article titled "A Roadmap for Big Model," published on Arxiv's preprint website on March 26 this year, was suspected of serious plagiarism.
The article is a 200-page academic review paper jointly completed by many universities and enterprises at home and abroad, with as many as 100 authors, from domestic universities such as Tsinghua University and Peking University, foreign universities such as Columbia University and University of Montreal, byteDance, Huawei, JD.com, Tencent and other enterprises, as well as institutions such as the Chinese Academy of Sciences and Beijing Zhiyuan.
In his article "A Case of Plagarism in Machine Learning Research," Nicholas Carlini detailed the suspicion of large sections of plagiarism of other papers, evidenced by large-scale text overlap, including his own paper, "Deduplicating Training Data Makes." Language Models Better”。
Nicholas Carlini explained in the article that they found many of the sentences very familiar when reading the paper, and then found more evidence by using the paper double-check tool, and then accurately located the text content suspected of plagiarism in the "Roadmap to the Big Model" paper and existing papers.
The paper found that the questioned paper has now been commented on the page by its publishing site Arxiv that there is a "text overlap" between the article and Nicholas Carlini's paper.
Previously, on March 31, the Beijing Zhiyuan community wrote an article entitled "How to Refine the Big Model?" 200 pages pdf100+ authors 19 units! Beijing Zhiyuan Tsinghua Tang Jie et al. published" to introduce the paper:
"With the rapid development of AI technology represented by deep learning, the training application mode of intelligent models has gradually changed from 'big refining model' to 'refining big model'. The study of large models has developed rapidly in recent years, and the amount of parameters of models has expanded at an alarming rate. The "A Roadmap for Big Model" recently released by Beijing Zhiyuan Artificial Intelligence Research Institute is led by Tang Jie, the leader of the Wudao Big Model Research Project, the academic vice dean of Zhiyuan, and the professor of the Department of Computer Science of Tsinghua University, starting from the four levels of basic resources of large models, large model construction, key technologies of large models and application exploration of large models, and comprehensively introduces and discusses 16 related topics in 15 specific fields. Very noteworthy. ”
The "big model", the research subject in the "Roadmap for the Big Model", is a hot area of concern for artificial intelligence in recent years. Artificial intelligence has developed to this day, and models with huge parameters such as GPT and BERT have been developed, and they have made unprecedented achievements in fields such as computer vision and natural language processing. At the same time, because of the huge number of parameters of large models, recently the academic community has begun to study them as a special type of artificial intelligence model.
In a paper published last year and signed by about 100 authors, led by Stanford professor Li Feifei and others, referring to such models as the "Foundation Model," Tsinghua's main participant focused on the future research path of the "Big Model" in the same sense. In the Tsinghua paper, the authors describe that "the mentioned foundation Chinese model is known as the BMs, and it is called Big Model in the Chinese context."
The "suspected plagiarism" incident caused a heated discussion on foreign social media. Gautam Kamath, a professor at the University of Waterloo, said he was surprised that for an article with so many authors, none of them noticed the similarity and corrected it.