Chinese scholars report the IBM team plagiarism, and the relevant party responded: the investigation has been launched

Author:Scientific network Time:2022.06.30

Text | "China Science News" reporter Zhao Guangli, a trainee reporter Tian Ruiying

Due to the successful plagiarism incident, the top meeting in the field of computer vision -CVPR 22 (2022 International Computer Visual and Mode Recognition Conference) has recently been circled.

On June 24, those who broke the news played the author and posted a funny video on the social platform. "Self -exposure" had more than a dozen articles. Some plagiarism paragraphs were not bad. Essence This passage of God has set off a bottom of the AI research team of this CVPR 22 outstanding papers, the AI Research Team of Seoul University.

Immediately after June 25, Chinese scholar Qi Xianbiao posted in the online community, accusing the IBM Zurich Research Institute a research team plagiarized by a paper accepted by the CVPR 22 conference. Idea. Ironically, the organizer of this competition is IBM itself.

The serious academic top will change the "melon -eating scene" in seconds. The amount of submission is large, the review is not strict, the ethical review mechanism is not complete, etc. It may be the reason. However, insiders in the computer vision circle said in an interview with the China Science News that the plagiarism incident was revealed that "just a microcosm of the CV circle 'inner roll" "and lamented that" a field is too hot, it may not be a good thing. "

Chinese scholars report foreign team "plagiarism IDEA"

Qi Xianbiao, who is currently working on a innovative research institution in Southern Southern, found that his papers were plagiarized by his friends' reminders. This plagiarism paper has been accepted and published by CVPR 22.

On June 25, he posted a post in Zhihu, pointing directly at a research team of IBM Zurich Research Institute, and plagiarizing he participated in the ICDAR 2021 in the Ping An VC Group team in 2021 (document analysis and identification international conference, which is the field of optical text recognition fields The results of the flagship international conference) contest.

Qi Xianbiao believes that the IBM research team copied their formalities, including methodology, network structure, data processing methods (including pre -processing and post -processing), source code, visualization methods, and several tips in their implementation. Almost "the entire complex system", but did not make any labels in the literature quotation.

Qi Xianbiao revealed the plagiarism of the IBM Zurich Research Institute

"This is simply a joke. We participated in the ICDAR2021 competition hosted by IBM, but our papers were plagiarized by another group of IBM employees." Qi Xianbiao said in an exclusive interview with the China Science News that last year they were for the ICDAR 2021 competition. The team members fought at the company for seven or eight weekends. "The sweat we paid should not be wiped out by these plagiarists."

Qi Xianbiao told reporters that in the first half of last year, he led the team to participate in five competitions in the event, and the plagiarized papers won the second place in the track identification track. After that, they form the papers in June and upload it to the ARXIV platform.

However, Qi Xianbiao believed that the other party's plagiarism was very "clever": "They did not copy in the paper text, but plagiarized our entire IDEA, and packed them in their ways.

The plagiarism team is Ahmed Nassar, Nikolaos Livathinos, Maksym Lysak and Peter Staar from the IBM Zurich Research Institute. Qi Xianbiao said: "They eat the most essence of IDEA, and then send the article to the CVPR conference's paper set."

Under the anger, Qi Xianbiao wrote a report email to the chairman of the CVPR Conference Programming Committee and the vice president of IBM responsible for artificial intelligence, and published the original committee of the incident on the Reddit website and Zhihu, which caused a stir.

The popularity of network fermentation exceeds the expectations of Qixian. However, he bluntly said that "not named nomads and no fight", "just want to be justified academically."

"If you can send someone else's papers on the CVPR, what innovation do we still do hard?" Open source code is used to help the field develop and benefit the community, not for these plagiarismers to pollute the entire academic circle. "

At the end of the email to the chairman of the CVPR conference, Qi Xianbiao wrote: "Considering so many plagiarism, I call on each top meeting to set up an ethical committee to do ethical review of academic dislocation."

For Qi Xianbiao's report, some people think that it is indeed a "real hammer plagiarism", and some people think it is best to "let the bullet fly for a while."

"I made this direction in a form recognition, and I suspected it was copied at first glance." In the post of Qixian's target, a colleagues who pseudo the pseudonym "inner roll" said: After, the code has been seen from beginning to end. "Personal opinions can be said to be hammering.

After reading the report letter and proof of Qi Xianbiao, a technical expert from an online education technology company told the China Science News that "the possibility of plagiarism is very high, but it is not easy to make conclusion now." He said that this matter Now that you have complained to the CVPR Organizing Committee, it is best to wait for the regular program to confirm the results. Another interviewed experts believe that in the current hot CV circle, the boundary of plagiarism and reference is also very vague. If it is not meticulous in the same direction, it may be difficult to identify.

One of South Korea's self -conscious plagiarism, teammates have "dumped the pot"

Coincidentally, at this CVPR 22 meeting, another plagiarism incident in the papers stirred the academic circle.

On June 24th, netizens named "E2V-SDE (Parody)" posted a video of 7 minutes and 16 seconds on YouTube. "Self-explosion" has been "sutured" by plagiarism by CVPR 22. Specific plagiarism clips. Soon after the video was released, the playback volume exceeded 100,000.

The papers accused of plagiarism have also been selected as ORAL (oral report papers), the top 4%of the top 4%of the CVPR, and the 6 authors of the papers are from Seoul University and the Korean Institute of Science and Technology.

It is worth noting that Yin Shenglai, a professor of the AI Research Institute of Seoul University, as the author of the paper, is an authoritative expert in the AI field in South Korea. He has served as chairman of the Fourth Industrial Revolutionary Committee directly under the President of Korea. The department was selected as "Chief Researcher"; another author of the paper was also stated by netizens that the children of the Ministry of Science and Technology Information and Communications (Ministry of Science and Technology) Li Zonghao.

The clips accused of plagiarism involved different chapters such as Method, IntRoduction, and Preliminaries. Take the dissertation of the University of Toronto, Canada in 2019 as an example. Some of the papers published by the Seoul University team are even completely copied.

The person who broke the news then posted the exposure video on Twitter, and the author who had been copied one by one to show "thanks".

Figure: "Thanks" by the news

Within a day, the CVPR official responded that the papers at Seoul University had submitted the Society of Electrical Electronics Engineers (IEEE) for investigation and withdrew the paper.

Jongwan Kim, the first author of the paper, also responded to Twitter that the errors of the thesis were caused by myself. "I agree that plagiarism is absolutely unforgivable. I will accept any discipline without any excuses. "

The first author of the paper Jin Zhongwan responded

After the incident was exposed, the other authors of the thesis had protected themselves and "shaken the pot."

The co -author of the paper responded with Byunggook Na with posts: "I apologize for this absurd thing. As a thesis collaborator, I should check the plagiarism of the paper. Although I have not thought of the plagiarism of the first author at all, I deeply feel that I have I have myself. responsibility."

Yin Shenglun also responded through Twitter. He said that he did not know the plagiarism of the paper before. "Under any circumstances, plagiarism should not be allowed. I don't know how to apologize for this case."

People do not buy it for teammates' "shaking pot" behavior.

"This kind of thing happened, and the tutor of the paper did not modify the paper." Wang Yichao, an expert in Alida Academy algorithm, felt that the disserted collaborators did not understand or even read the paper.

Under the scandal, the domestic academic circle in South Korea boiled anti -Yingtian. Many Korean researchers have spoken through various channels to condemn "behaviors like garbage, pour cold water for other researchers", and demand that "relevant units can deal with this matter seriously." In the Korean public opinion field, people are worried that the incident may be fermented into the second "Huang Yuxi incident".

IEEE response: The investigation has been started

Different from the Korean AI team plagiarism incident, it is basically concluded, and the report of Qi Xianbiao needs to wait for the results.

Qi Xianbiao told the Journal of China that on the morning of June 30, the chairman of the procedure committee of the CVPR Conference wrote to them, stating that they had been concerned about the reason he reported, and had transferred suspected plagiarism papers to IEEE for further investigation.

Jennifer Carruth, the content quality guarantee manager of the IEEE Computer Association, also subsequently wrote to him, saying that "the allegations have been conducted on this investigation."

"I was told‘ Once I have the results of the survey, I will tell you ’, but this survey will definitely take a while." Qi Xianbiao told reporters. He also said that as of now, the IBM and IBM Zurich Research Institute has not responded any response.

The plagiarism incident attracted the ridicule of netizens, and the prestigious Top CVPR became "CTR+C, CTR+V, Plagiarism, Replication" (copy, paste, plagiarism, and repeat). Someone couldn't help asking: Didn't CVPR check it seriously?

"CVPR is too well -known, and now the submission volume is too large. There are any submitters, and the level of reviewers is uneven." Wang Yichao told the China Science Journal: "I feel that I feel academic ethics. I hope that through these things, it can be more standardized, and it is a good thing to expose. "

Digital shows that the amount of submission of the CVPR conference in 2022 reached 8,161, with a total of 2064 papers received, with a receiving rate of 25.28%; in the acceptable papers, 342 were selected as an ORAL paper. "CVPR ORAL plagiarism is just a microcosm of the computer vision (CV) circle." Liu Xin, technical experts in the field of automatic identification and data capture, and the executive director of Vision Cloud, said, looking at the CV circle, Idea collided with each other, Idea mutual collision, There are many things that are similar and moving flowers, but most of them have not been exposed. "It can be seen that a field is too hot, which may not be a good thing."

Related thesis information:

https://arxiv.org/abs/2206.07578

https://arxiv.org/pdf/2105.01848.pdf

Edit | Fangyuan

Capture | Zhihai

Data-nickName = "Science Network" Data-ALIAS = "ScienceNet-Cas" Data-Signature = "In-circle major events, industry insights, occasionally gossip ... Produced from the world's largest Chinese science community, Chinese Science News Agency. Contact us: Cooperation: Cooperation: Cooperation:[email protected]; submitting [email protected]. "Data-from =" 1 " /> Cooperation matters: [email protected]

Submission: [email protected]