雷鋒網(wǎng) AI 科技評(píng)論按：人工智能領(lǐng)域頂級(jí)學(xué)術(shù)會(huì)議之一的 IJCAI 2019 近日發(fā)布了論文錄用名單并通知了作者們。IJCAI 2019 今年一共收到 4752 篇有效提交論文，最終的論文收錄數(shù)量為 850 篇，接收率為 17.9%。

今年 IJCAI 的投稿總數(shù)比去年增加了約 37%，所以錄用率隨著降低到了不足 20% 也算情有可原。不過(guò)意外的是，收到錄用或者拒稿通知后，許多作者都通過(guò)種種途徑表達(dá)了自己的不滿，包括對(duì)于打分、審稿人意見、Rebuttal 環(huán)節(jié)等等，似乎都有大量問(wèn)題。一時(shí)間對(duì)于論文審稿過(guò)程的批評(píng)和討論沸沸揚(yáng)揚(yáng)，其它媒體也進(jìn)行了報(bào)道。

對(duì)于這樣的輿情，IJCAI 2019 的 PC （program chair，程序主席）Sarit Kraus 給作者們發(fā)送了一封郵件，做了一些解釋，也給出了一些他的個(gè)人觀點(diǎn)，希望可以平息作者們的憤懣。雷鋒網(wǎng) AI 科技評(píng)論摘錄并翻譯如下。

When I sent the notification messages on Friday, I was happy to end more than overwhelming six months of working on the reviewing process of IJCAI 2019. On the other hand I was very sad since while I made the authors of 850 papers very happy, I made the authors of 3902 papers very upset and annoyed.  Even though I have many papers accepted to highly refereed conferences, I believe I received many more rejection messages than most of you. As a result, any complaint you have about the reviewing process, its unfairness, its randomness, I am sure I thought the same over the years. I am grateful to all the authors that haven’t sent me a message to complain, and I will try answering the messages of the few that did. From first glance, it will probably not change the decisions.

當(dāng)我周五把所有的審稿結(jié)果都發(fā)送出去以后，IJCAI 2019 的長(zhǎng)達(dá)六個(gè)月的繁忙的審稿過(guò)程終于開心地結(jié)束了。另一方面，雖然我讓 850 篇論文的作者們很開心，但另外 3902 篇論文的作者們就一點(diǎn)也不開心，甚至有點(diǎn)煩躁；這也讓我不怎么好受。雖然我自己也曾經(jīng)有多篇論文被聲名顯赫的會(huì)議錄用，但我相信我曾經(jīng)收到過(guò)的拒稿通知也要比你們中大多數(shù)人更多。所以，大家對(duì)于審稿過(guò)程、其中的不公平、隨機(jī)性等等的各種報(bào)怨，我相信我在過(guò)去的這么多年中也有和你們一樣的感受。我很感謝所有沒(méi)發(fā)出報(bào)怨之聲的作者們，我也會(huì)盡量回復(fù)那些提出了質(zhì)疑的聲音。不過(guò)目前看來(lái)，這可能并不會(huì)改變審稿結(jié)果。

I tried hard to reduce the randomness of the process by decreasing the number of papers each PC and SPC received, by increasing the reviewing time period, and by trying to recruit good ACs/SPCs/PCs and reviewers. Almost all of them did a great job and I am very grateful. However, with 4752 submitted papers, it is extremely difficult. I believe we need to think and being innovative about the concept of highly refereed conferences.

為了減少審稿過(guò)程的隨機(jī)性我做了很多努力，減少了每位 PC 和 SPC（Senior PC）審的論文數(shù)量，增加了審稿時(shí)間，盡量找了好的 AC（Area Chair，領(lǐng)域主席）/SPC/PC 以及審稿人。幾乎所有人都做得很棒，我也非常欣慰。不過(guò)，我們畢竟收到了 4752 篇論文投稿，不管怎么樣這都是很難的一件事。我覺(jué)得我們需要重新思考“高被引頂會(huì)”這件事并有所創(chuàng)新。

I strongly believe that conferences are events for hearing new ideas and meeting other researchers and discussing research. My best projects started at conferences. I don’t think they are currently good for evaluations for getting promotion and funding.  We can think about making our journals better. Currently we have a journal track—each paper accepted to AIJ or JAIR with no prior publication in a conference is invited to be presented in IJCAI. We can further limit the number of papers each person can submit to IJCAI or any other innovative ways you all can come up with. We all need to think about it.

我堅(jiān)定認(rèn)為會(huì)議是聆聽新的想法、會(huì)見其它的研究者、討論學(xué)術(shù)科研的場(chǎng)所，我自己的最好的項(xiàng)目就是從一次會(huì)議開始的。但是頂會(huì)目前對(duì)于獲得升職、獲得經(jīng)費(fèi)并不能起到什么幫助。我們可以考慮把期刊做得更好。目前我們有一個(gè)期刊 track，每篇 AIJ 或者 JAIR 期刊接收的論文，如果此前沒(méi)有在會(huì)議中出現(xiàn)的，都會(huì)受邀在 IJCAI 展示。我們甚至可以考慮限制每位作者投稿到 IJCAI 的論文的數(shù)量，大家也可以一起想想其他的更創(chuàng)新的方式。我們需要共同思考。

In summary, I hope that all of you will come to IJCAI regardless of whether you have papers accepted or not. You can still submit papers to the IJCAI workshops or just come to meet friends and hear interesting and exciting talks.

總的來(lái)說(shuō)，不管你的論文有沒(méi)有被接收，我都希望大家全都可以來(lái)到 IJCAI 會(huì)議現(xiàn)場(chǎng)。你仍然可以把論文投稿到 IJCAI 的各個(gè) workshop，或者就是來(lái)見一見朋友、聽一聽有意思的演講。

Best regards,

Sarit Kraus

IJCAI 2019 Program Chair

對(duì)于這封郵件，有學(xué)者認(rèn)為觀點(diǎn)比較中肯，可以接受。雷鋒網(wǎng) AI 科技評(píng)論也認(rèn)為，雖然會(huì)議在計(jì)算機(jī)科學(xué)領(lǐng)域是非常重要的學(xué)術(shù)成果評(píng)議平臺(tái)，但是大批量往頂會(huì)投論文“灌水”的做法也已經(jīng)帶來(lái)了碩士生博士生申請(qǐng)門檻越來(lái)越高、參會(huì)門票難以購(gòu)買、審稿人年輕化等問(wèn)題，這都是需要領(lǐng)域內(nèi)的學(xué)者們反思并設(shè)計(jì)改進(jìn)方案的。

另外，AI 研習(xí)社近期有「頂會(huì)贊助計(jì)劃」。參與 IJCAI 2019 頂會(huì)討論小組活動(dòng)，積累研值排行TOP 3，將獲得 AI 研習(xí)社贊助的「直達(dá)頂會(huì)」的機(jī)票+酒店等參會(huì)費(fèi)用，讓你親臨大會(huì)現(xiàn)場(chǎng)，和大咖面對(duì)面~ 更多詳情請(qǐng)查看小組頁(yè)面 。

雷峰網(wǎng)原創(chuàng)文章，未經(jīng)授權(quán)禁止轉(zhuǎn)載。詳情見轉(zhuǎn)載須知。