{ "id": 381321, "title": "About my team's result and activities", "url": "/competitions/otto-recommender-system/discussion/381321", "author_name": "Makotu", "post_date": "2023-01-26T05:14:27.474000", "votes": 128, "comment_count": 80, "views": 0, "content": "

I know many of you are curious, so let me explain the current status of our team.

\n

・First of all, our team score is a merge of the 4 of us. We can release the results, including the code, after the competition is over.

\n

・Because there was a fear that specific person were not participating on their own, We asked him to share the code they had written as of yesterday morning. He said he achieved 0.596 himself, but did not share that code(for now, We don't know how he ended up with that result.) As a result, he was not able to share it with us, so we decided not be cooperating with that person. Specifically, we(4 of us) move other slack, and we do not have access to each other's information.

\n

・Of course, we do not share the code/data to him.

\n

I hope you understand that this is all we can do at this point, and that at this point we are basically four people thinking and working together.

\n

It is really sad to see such allegations in a competition for model accuracy. I too think that not only the prize winners, but also those who enter the gold zone should be provided with models and codes that produce the same results.

\n

I leave the decision to kaggle Team.
\nWe will be happy to answer any questions you may have.

\n

PS：
\nTo prevent this from happening in the future, slack communication with him will be open to public.

\n

\"\"

\n

\"\"

\n

\"\"

", "messages": [ { "id": 2115966, "postDate": "2023-01-26T05:14:27.473Z", "content": "

I know many of you are curious, so let me explain the current status of our team.

\n

・First of all, our team score is a merge of the 4 of us. We can release the results, including the code, after the competition is over.

\n

・Because there was a fear that specific person were not participating on their own, We asked him to share the code they had written as of yesterday morning. He said he achieved 0.596 himself, but did not share that code(for now, We don't know how he ended up with that result.) As a result, he was not able to share it with us, so we decided not be cooperating with that person. Specifically, we(4 of us) move other slack, and we do not have access to each other's information.

\n

・Of course, we do not share the code/data to him.

\n

I hope you understand that this is all we can do at this point, and that at this point we are basically four people thinking and working together.

\n

It is really sad to see such allegations in a competition for model accuracy. I too think that not only the prize winners, but also those who enter the gold zone should be provided with models and codes that produce the same results.

\n

I leave the decision to kaggle Team.
\nWe will be happy to answer any questions you may have.

\n

PS：
\nTo prevent this from happening in the future, slack communication with him will be open to public.

\n

\"\"

\n

\"\"

\n

\"\"

", "rawMarkdown": "I know many of you are curious, so let me explain the current status of our team.\n\n・First of all, our team score is a merge of the 4 of us. We can release the results, including the code, after the competition is over.\n\n・Because there was a fear that specific person were not participating on their own, We asked him to share the code they had written as of yesterday morning. He said he achieved 0.596 himself, but did not share that code(for now, We don't know how he ended up with that result.) As a result, he was not able to share it with us, so we decided not be cooperating with that person. Specifically, we(4 of us) move other slack, and we do not have access to each other's information.\n\n・Of course, we do not share the code/data to him.\n\nI hope you understand that this is all we can do at this point, and that at this point we are basically four people thinking and working together.\n\nIt is really sad to see such allegations in a competition for model accuracy. I too think that not only the prize winners, but also those who enter the gold zone should be provided with models and codes that produce the same results.\n\nI leave the decision to kaggle Team. \nWe will be happy to answer any questions you may have.\n\nPS：\nTo prevent this from happening in the future, slack communication with him will be open to public. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F397595%2F4247c019da1b917bd7d18569426fd2c9%2Fslack_1.png?generation=1674712574709325&alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F397595%2F2bd07680a672bfbcff0f156af9a8474c%2Fslack_v2.png?generation=1674712590241405&alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F397595%2F95628b01fdc797181330a9e83d82d1f3%2Fslack_3.png?generation=1674712604829685&alt=media)", "votes": 127 }, { "id": 2119716, "postDate": "2023-01-29T03:33:38.697Z", "content": "

The sad part is that getting a gold medal in a Kaggle competition is very hard. Getting a solo gold is extremely hard. How many honest people were denied gold medals in the past because of cheating? People who spent countless hours and nights training their models…

\n

The very existence of such schemes destroys the fun and competitive aspect of Kaggle competitions. Kaggle should treat those matters extremely seriously if they want to avoid the dilution of the Kaggle Competitions brand, especially the competitions Grandmaster brand, which now stands for quite a lot (there are companies recruiting based on obtaining the GM title).

", "rawMarkdown": "The sad part is that getting a gold medal in a Kaggle competition is very hard. Getting a solo gold is extremely hard. How many honest people were denied gold medals in the past because of cheating? People who spent countless hours and nights training their models...\n\nThe very existence of such schemes destroys the fun and competitive aspect of Kaggle competitions. Kaggle should treat those matters extremely seriously if they want to avoid the dilution of the Kaggle Competitions brand, especially the competitions Grandmaster brand, which now stands for quite a lot (there are companies recruiting based on obtaining the GM title).", "votes": 35, "replies": [ { "id": 2120307, "postDate": "2023-01-29T13:59:44.407Z", "content": "

Thanks for your words. Being a GM is on my before-30 bucket list but sadly I got 2 silvers last year:(. I still have 9 months to achieve this haha, I think I will stick with code competition…

", "rawMarkdown": "Thanks for your words. Being a GM is on my before-30 bucket list but sadly I got 2 silvers last year:(. I still have 9 months to achieve this haha, I think I will stick with code competition...", "votes": 4, "replies": [ { "id": 2120960, "postDate": "2023-01-29T23:21:51.653Z", "content": "

Keeping my fingers crossed for you. I still remember your starting at the HomeCredit competition.

", "rawMarkdown": "Keeping my fingers crossed for you. I still remember your starting at the HomeCredit competition.", "votes": 1 } ] } ] }, { "id": 2117522, "postDate": "2023-01-27T10:32:31.633Z", "content": "

I'm just wondering why he deleted his name from his github yesterday.
\n
\nhttps://github.com/poteman/poteman/commit/de214efb1495151385fc199dd8d885541fabdf9e

", "rawMarkdown": "I'm just wondering why he deleted his name from his github yesterday.\n\nhttps://github.com/poteman/poteman/commit/de214efb1495151385fc199dd8d885541fabdf9e ", "votes": 31, "replies": [ { "id": 2117716, "postDate": "2023-01-27T13:41:43.060Z", "content": "

@poteman Now we can't see your name. Well done😄👍
\n

", "rawMarkdown": "@poteman Now we can't see your name. Well done😄👍\n", "votes": 15, "replies": [ { "id": 2118433, "postDate": "2023-01-28T02:27:53.297Z", "content": "

@poteman We can still develop your \"Hi there\" OSS. This is the beauty of OSS.
\n
\nhttps://github.com/yanqiangmiffy/poteman

", "rawMarkdown": "@poteman We can still develop your \"Hi there\" OSS. This is the beauty of OSS.\n\nhttps://github.com/yanqiangmiffy/poteman", "votes": 19 } ] } ] }, { "id": 2116033, "postDate": "2023-01-26T06:24:31.360Z", "content": "

I teamed up with him in the past PANDA competition.
\nHe said he could not share his training code due to a paper submission, and we thought that was ok since our private 1st sub did not contain his model. So we left out his codes in the summary and the winner's call.

\n

Looking at this incident, I looked back to his notebook and found a strange match between his code and the Public 3rd, Private 20th team's solution. (https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/169437)

\n

This model (MultiheadAttentionClassifier) is quite unique to my knowledge, and I could not find other code match after searching in the github. Only SHUJUN's codes appear after the search, and it wasn't public during the competition was held. (There might be more than 2 suspects?)

\n

\"\"

", "rawMarkdown": "I teamed up with him in the past PANDA competition.\nHe said he could not share his training code due to a paper submission, and we thought that was ok since our private 1st sub did not contain his model. So we left out his codes in the summary and the winner's call.\n\nLooking at this incident, I looked back to his notebook and found a strange match between his code and the Public 3rd, Private 20th team's solution. (https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/169437)\n\nThis model (MultiheadAttentionClassifier) is quite unique to my knowledge, and I could not find other code match after searching in the github. Only SHUJUN's codes appear after the search, and it wasn't public during the competition was held. (There might be more than 2 suspects?)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3508221%2F6bc0a5590056ca45b967bb544bf9e571%2F2023-01-26%20152028.png?generation=1674714058896654&alt=media)", "votes": 29, "replies": [ { "id": 2116044, "postDate": "2023-01-26T06:37:15.517Z", "content": "

The network part is quite identical as well..

\n

https://github.com/Shujun-He/PANDA/blob/e0327b1a604b4c1444827262c931daa47c10855c/pandalayer1test8_s64/Network.py#L83

", "rawMarkdown": "The network part is quite identical as well..\n\nhttps://github.com/Shujun-He/PANDA/blob/e0327b1a604b4c1444827262c931daa47c10855c/pandalayer1test8_s64/Network.py#L83", "votes": 2 }, { "id": 2116046, "postDate": "2023-01-26T06:40:20.427Z", "content": "

Or he didn't even have code at that time. After the competition ended, he copied someone elses solution in order to drop suspicions.

", "rawMarkdown": "Or he didn't even have code at that time. After the competition ended, he copied someone elses solution in order to drop suspicions.", "votes": 3, "replies": [ { "id": 2116047, "postDate": "2023-01-26T06:42:09.640Z", "content": "

No, this was a notebook competition and the above are his codes* during* the competition was held.

", "rawMarkdown": "No, this was a notebook competition and the above are his codes* during* the competition was held.", "votes": 2 } ] }, { "id": 2118580, "postDate": "2023-01-28T06:02:24.977Z", "content": "

Thanks for bringing this to light @kyoshioka47 . It is indeed very strange that @poteman had the exact same code as my solution in PANDA. In fact, my account was (wrongly) banned as a result of a major compliance action following SIIM-ISIC Melanoma Classification, because Kaggle thought I sold my code to a member of angtk. You can check out @group16 's excellent investigative work about angtk here: https://www.kaggle.com/competitions/siim-isic-melanoma-classification/discussion/175156.

\n

I knew for a fact that I did not have anything to do with angtk, so I immediately appealed to Kaggle to get my account restored. Keep in mind that my real identity on Linkedin is linked to my Kaggle account so you can see how serious that was for me: effectively I was labeled a cheater and I needed to clear my name. Luckily, after a few email exchanges, I was able to convince Kaggle to restore my account shortly after. I think Kaggle did not have clear evidence and only linked me with angtk since I released my code on github.

\n

After speaking to @group16 , we think that @ruozha001 was connected with @poteman and sold my code to @poteman after she got her hands on it. You can also check out the post made by @group16 (https://www.kaggle.com/competitions/otto-recommender-system/discussion/381588).

", "rawMarkdown": "Thanks for bringing this to light @kyoshioka47 . It is indeed very strange that @poteman had the exact same code as my solution in PANDA. In fact, my account was (wrongly) banned as a result of a major compliance action following SIIM-ISIC Melanoma Classification, because Kaggle thought I sold my code to a member of angtk. You can check out @group16 's excellent investigative work about angtk here: https://www.kaggle.com/competitions/siim-isic-melanoma-classification/discussion/175156.\n\nI knew for a fact that I did not have anything to do with angtk, so I immediately appealed to Kaggle to get my account restored. Keep in mind that my real identity on Linkedin is linked to my Kaggle account so you can see how serious that was for me: effectively I was labeled a cheater and I needed to clear my name. Luckily, after a few email exchanges, I was able to convince Kaggle to restore my account shortly after. I think Kaggle did not have clear evidence and only linked me with angtk since I released my code on github. \n\nAfter speaking to @group16 , we think that @ruozha001 was connected with @poteman and sold my code to @poteman after she got her hands on it. You can also check out the post made by @group16 (https://www.kaggle.com/competitions/otto-recommender-system/discussion/381588). ", "votes": 21, "replies": [ { "id": 2119123, "postDate": "2023-01-28T14:01:01.487Z", "content": "

I am afraid to merge with anyone who I don't know now 🤯

", "rawMarkdown": "I am afraid to merge with anyone who I don't know now 🤯" } ] }, { "id": 2119573, "postDate": "2023-01-28T22:41:34.703Z", "content": "

This competition was held 3 years ago and I have forgotten many details.

\n
    \n
  1. What evidence is there that this part of the code belongs to me? Especially I already have the codes before the game ends.
    2. \n
  2. If the first point is true, what evidence is there to prove that the codes are not written by myself?
    3. \n
", "rawMarkdown": "This competition was held 3 years ago and I have forgotten many details.\n1. What evidence is there that this part of the code belongs to me? Especially I already have the codes before the game ends.\n2. If the first point is true, what evidence is there to prove that the codes are not written by myself?", "votes": -10, "replies": [ { "id": 2119632, "postDate": "2023-01-29T00:23:56.587Z", "content": "

@poteman Hi there, I'm Kazuki Onodera👋

\n
\n

I already have the codes before the game ends.

\n
\n

So you're saying this is a coincidence. I think this is bad move. You should have said \"I copied his code after this comp\".

", "rawMarkdown": "@poteman Hi there, I'm Kazuki Onodera👋\n> I already have the codes before the game ends.\n\nSo you're saying this is a coincidence. I think this is bad move. You should have said \"I copied his code after this comp\"." }, { "id": 2119668, "postDate": "2023-01-29T01:40:23.223Z", "content": "

He may be saying : \"What evidence is there .. that I already have the codes before the games ends?\"

\n

I think folks are getting silly, maybe take a beat and consider what are reasonable assumptions and what might be a bit unfair.

\n

Tracking the provenance of code is not a simple task that can be done off the cuff. It requires an immense amount of effort. I've found things on git only after a great deal of digging, it's very deep web and not something that will necessarily show up on a one off google search.

\n

Things also get deleted but still stay in personal repos. There can be all sorts of explanations.

", "rawMarkdown": "He may be saying : \"What evidence is there .. that I already have the codes before the games ends?\"\n\nI think folks are getting silly, maybe take a beat and consider what are reasonable assumptions and what might be a bit unfair.\n\nTracking the provenance of code is not a simple task that can be done off the cuff. It requires an immense amount of effort. I've found things on git only after a great deal of digging, it's very deep web and not something that will necessarily show up on a one off google search.\n\nThings also get deleted but still stay in personal repos. There can be all sorts of explanations." }, { "id": 2119763, "postDate": "2023-01-29T04:47:19.590Z", "content": "

Well, his codes are submitted during the completion and the Kaggle team can easily check this. If the team can mail me about this, I am happy to point out.

", "rawMarkdown": "Well, his codes are submitted during the completion and the Kaggle team can easily check this. If the team can mail me about this, I am happy to point out.", "votes": 1 }, { "id": 2119802, "postDate": "2023-01-29T05:46:43.893Z", "content": "
\n

This competition was held 3 years ago and I have forgotten many details.

\n
    \n
  1. What evidence is there that this part of the code belongs to me? Especially I already have the codes before the game ends.
    2. \n
  2. If the first point is true, what evidence is there to prove that the codes are not written by myself?
    3. \n
\n
\n
    \n
  1. It's on your github… 🤦‍♂️🍩
    2. \n
  2. While not being caught red-handed, the probability of sharing so much similarities (it's the exact same) between two pieces of code is astronomically small, except when this is taken from a public resource (but this is not the case). Both codes even contain no spaces before & after the =. Your public notebooks on Kaggle, however, seem to always include spaces?
    3. \n
", "rawMarkdown": "> This competition was held 3 years ago and I have forgotten many details.\n> 1. What evidence is there that this part of the code belongs to me? Especially I already have the codes before the game ends.\n> 2. If the first point is true, what evidence is there to prove that the codes are not written by myself?\n\n1. It's on your github... 🤦‍♂️🍩\n2. While not being caught red-handed, the probability of sharing so much similarities (it's the exact same) between two pieces of code is astronomically small, except when this is taken from a public resource (but this is not the case). Both codes even contain no spaces before & after the `=`. Your public notebooks on Kaggle, however, seem to always include spaces?\n\n", "votes": 4 } ] }, { "id": 2119833, "postDate": "2023-01-29T06:18:57.313Z", "content": "

Oh, for some reason you seem to have deleted all your submission notebooks for panda competition.

\n

\"\"

\n

This will reduce the evidence to prove your innocence.

\n

But don't worry. The code is still on the kaggle server because it has been submitted. Let's check it with kaggle staff.
\nAt the same time, comparing it to @shujun717 's subs will prove your innocence. Good for you👍

", "rawMarkdown": "Oh, for some reason you seem to have deleted all your submission notebooks for panda competition.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1670024%2F5d565cf58ff18f86d50971b8df1533c4%2F2023-01-29%2015.10.36.png?generation=1674972707288402&alt=media)\n\nThis will reduce the evidence to prove your innocence.\n\nBut don't worry. The code is still on the kaggle server because it has been submitted. Let's check it with kaggle staff.\nAt the same time, comparing it to @shujun717 's subs will prove your innocence. Good for you👍", "votes": 12 }, { "id": 2122932, "postDate": "2023-01-31T08:24:33.340Z", "content": "

@kyoshioka47 @yukkyo @shujun717 I think i got it, what happend in PANDA. For some time i cant wrap my head around why there are fake teams above poteman on public:

\n
    \n
  1. poteman joins with K and Y
    2. \n
  2. ruozha001 joins S. Clearly its a fake \"female\" account to target teams that can do well and leak results afterwards
    3. \n
  3. S goes high on public, ruozha001 leaks exact ensemble to other \"boost\" teams, they are fake-users only
    4. \n
  4. \"boost\" teams try to worse the S score and post it, you can still see some of those teams on LB, not banned
    5. \n
  5. poteman at the same time is stuck with K and Y, they are 20-ish on public. He understands that stealing whole S ensemble is super-risky, as he has K&Y in his team, they are not fake
    6. \n
  6. poteman steals just LB best single model from S ensemble, its around .910, thats all he can do, he is upset at 20th place with those \"losers\" K&Y
    7. \n
  7. K&Y just win competition without poteman stolen model.
    8. \n
\n

If all of that true, then S should have best single model around .910 LB, bb being resnext50-whatever it was.

", "rawMarkdown": "@kyoshioka47 @yukkyo @shujun717 I think i got it, what happend in PANDA. For some time i cant wrap my head around why there are fake teams above poteman on public:\n1. poteman joins with K and Y\n2. ruozha001 joins S. Clearly its a fake \"female\" account to target teams that can do well and leak results afterwards\n3. S goes high on public, ruozha001 leaks exact ensemble to other \"boost\" teams, they are fake-users only\n4. \"boost\" teams try to worse the S score and post it, you can still see some of those teams on LB, not banned\n5. poteman at the same time is stuck with K and Y, they are 20-ish on public. He understands that stealing whole S ensemble is super-risky, as he has K&Y in his team, they are not fake\n6. poteman steals just LB best single model from S ensemble, its around .910, thats all he can do, he is upset at 20th place with those \"losers\" K&Y\n7. K&Y just win competition without poteman stolen model.\n\nIf all of that true, then S should have best single model around .910 LB, bb being resnext50-whatever it was. \n", "votes": 5 } ] }, { "id": 2116011, "postDate": "2023-01-26T06:08:31.477Z", "content": "

In another competition, I have received a team merge request from two of alleged kagglers at the almost same time. I was in the gold zone at the time and got an email from the master first.

\n

He had never submitted at the competition but had a great record in the past competitions and we were considering merging, but we decided not to because we considered it a risk to invite a stranger who had never submitted.
\nShortly thereafter, we received a merge request from a grandmaster. Surprisingly, with only a few exceptions, the email was almost identical to the email from the master in terms of the position of the surprise mark and the wording of the text.

\n

I suspected that there might be some connection between the two kagglers. Then I discovered that they had both performed well together quite often, even though they had been on different teams in past competitions. I suspected some kind of private sharing or even the same person.

\n

I did not report this about it as it was only a circumstantial suspicion. However now, I strongly suspect that these two are in cahoots, or are the same person, and are acting to subvert the other team members to their own advantage.

\n

I have teamed up with makotu in the past. He was a very enthusiastic kaggler, quick to share good experimental code, and extremely talented. I posted this in the hope that my experience could reinforce makotu's assertion.

", "rawMarkdown": "In another competition, I have received a team merge request from two of alleged kagglers at the almost same time. I was in the gold zone at the time and got an email from the master first.\n\nHe had never submitted at the competition but had a great record in the past competitions and we were considering merging, but we decided not to because we considered it a risk to invite a stranger who had never submitted.\nShortly thereafter, we received a merge request from a grandmaster. Surprisingly, with only a few exceptions, the email was almost identical to the email from the master in terms of the position of the surprise mark and the wording of the text.\n\nI suspected that there might be some connection between the two kagglers. Then I discovered that they had both performed well together quite often, even though they had been on different teams in past competitions. I suspected some kind of private sharing or even the same person.\n\nI did not report this about it as it was only a circumstantial suspicion. However now, I strongly suspect that these two are in cahoots, or are the same person, and are acting to subvert the other team members to their own advantage.\n\nI have teamed up with makotu in the past. He was a very enthusiastic kaggler, quick to share good experimental code, and extremely talented. I posted this in the hope that my experience could reinforce makotu's assertion.", "votes": 17, "replies": [ { "id": 2116019, "postDate": "2023-01-26T06:16:00.700Z", "content": "

This mail is from the grand master

\n
\n

Hi, YYama.

\n

Hope you are doing well!
\n I am writing this to you to check if you are interested in teaming up with me for the competition \"RSNA 2022 Cervical Spine Fracture Detection\".
\n We can make a team together and get better score.

\n

Looking forward to your reply.

\n
\n

This mail is from the master

\n
\n

Hi, YYama
\n Hope you are doing well!
\n I am writing this to you to check if you are interested in teaming up with me for the competition RSNA

\n

About device:
\n 16 * A100 gpu and colob pro, and I can also rent gpu if needed.

\n

We can make a team together and get better score.
\n Looking forward to your reply.

\n
\n

They wouldn't match so much by coincidence. But if I had only received an email from one of them, I would not have suspected cheating.

", "rawMarkdown": "This mail is from the grand master\n\n> Hi, YYama.\n> \n> Hope you are doing well!\n> I am writing this to you to check if you are interested in teaming up with me for the competition \"RSNA 2022 Cervical Spine Fracture Detection\". \n> We can make a team together and get better score.\n> \n> Looking forward to your reply.\n\nThis mail is from the master\n\n> Hi, YYama\n> Hope you are doing well!\n> I am writing this to you to check if you are interested in teaming up with me for the competition RSNA\n> \n> About device:\n> 16 * A100 gpu and colob pro, and I can also rent gpu if needed. \n> \n> We can make a team together and get better score.\n> Looking forward to your reply.\n\nThey wouldn't match so much by coincidence. But if I had only received an email from one of them, I would not have suspected cheating.", "votes": 6, "replies": [ { "id": 2116024, "postDate": "2023-01-26T06:18:36.063Z", "content": "

Even I got the mails in close proximity for this competition. I am sure, if I teamed up with one, the other would not have mailed

", "rawMarkdown": "Even I got the mails in close proximity for this competition. I am sure, if I teamed up with one, the other would not have mailed", "votes": 2 }, { "id": 2116048, "postDate": "2023-01-26T06:42:20.247Z", "content": "

I received a similar email.

\n
Hi, zakopuro.\n\nI hope all is well!\nI'm writing this to you to check if you are interested in teaming up with me for the competition 'American Express'. \nWe can make a team together and get better score.\n\nLooking forward to your reply.\n
\n

When I teamed up with him in the past, he said his PC was broken and he couldn't share code.
\nAgain, was he unable to share the code because his PC was broken?😂

", "rawMarkdown": "I received a similar email.\n\n```\nHi, zakopuro.\n\nI hope all is well!\nI'm writing this to you to check if you are interested in teaming up with me for the competition 'American Express'. \nWe can make a team together and get better score.\n\nLooking forward to your reply.\n```\n\n\nWhen I teamed up with him in the past, he said his PC was broken and he couldn't share code.\nAgain, was he unable to share the code because his PC was broken?😂", "votes": 8 }, { "id": 2116163, "postDate": "2023-01-26T09:07:10.960Z", "content": "

Same here.

\n

In HPA+HUBMAP last year, our team also recieved the similar team-merge email from him. At that time, we were in gold zone and he did not even make a single submission!

\n

\"\"

", "rawMarkdown": "Same here. \n\nIn HPA+HUBMAP last year, our team also recieved the similar team-merge email from him. At that time, we were in gold zone and he **did not even make a single submission!**\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2909490%2Fd0241d13a0132fdfbfce75184380caaa%2F_20230126170540.jpg?generation=1674723958239883&alt=media)", "votes": 6 }, { "id": 2116249, "postDate": "2023-01-26T10:53:40.647Z", "content": "

I received a very similar Email from him in H&M competition two weeks before the submission deadline, when I was rankinng 3rd and he didn't even submit anything. \"\"

", "rawMarkdown": "I received a very similar Email from him in H&M competition two weeks before the submission deadline, when I was rankinng 3rd and he didn't even submit anything. ![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F317080%2Ff677953f17363269f2302e5bea946735%2F2023-01-26%2018.52.46.png?generation=1674730407537480&alt=media)", "votes": 2 }, { "id": 2116330, "postDate": "2023-01-26T12:27:03.883Z", "content": "

@carnozhao Just curious, can your team reproduce the best submission from scratch?

", "rawMarkdown": "@carnozhao Just curious, can your team reproduce the best submission from scratch?", "votes": 3 }, { "id": 2116337, "postDate": "2023-01-26T12:32:00.290Z", "content": "

Yes, if you ignore model training randomness

", "rawMarkdown": "Yes, if you ignore model training randomness" }, { "id": 2116347, "postDate": "2023-01-26T12:39:35.107Z", "content": "

That means you already checked reproducibility of your teammate codes?

", "rawMarkdown": "That means you already checked reproducibility of your teammate codes?" }, { "id": 2116352, "postDate": "2023-01-26T12:41:58.740Z", "content": "

yes, most of code is written by me

", "rawMarkdown": "yes, most of code is written by me", "votes": 2 }, { "id": 2116356, "postDate": "2023-01-26T12:45:59.913Z", "content": "

Thanks, I'm relieved to hear that. In this situation, we can't trust random guys's submission file.

", "rawMarkdown": "Thanks, I'm relieved to hear that. In this situation, we can't trust random guys's submission file.", "votes": 1 }, { "id": 2116362, "postDate": "2023-01-26T12:50:49.047Z", "content": "

what I am concerned is about, is it possible some people share the team's solution outside it ? I don't intend to blame anyone, but looks like this could be the case with these two coming up, so suddenly at the end of the competition. I think their first few submissions needs to be dug up, is it possible to identify the source. Someone within a high scoring team, probably shared their submisison with these 2 guys

", "rawMarkdown": "what I am concerned is about, is it possible some people share the team's solution outside it ? I don't intend to blame anyone, but looks like this could be the case with these two coming up, so suddenly at the end of the competition. I think their first few submissions needs to be dug up, is it possible to identify the source. Someone within a high scoring team, probably shared their submisison with these 2 guys", "votes": 3 }, { "id": 2116368, "postDate": "2023-01-26T12:55:24.507Z", "content": "

That's what I thought. We still don't know how they can always join the leaderboard in gold zone. Someone from top teams might be leaking their submissions.

", "rawMarkdown": "That's what I thought. We still don't know how they can always join the leaderboard in gold zone. Someone from top teams might be leaking their submissions.", "votes": 3 }, { "id": 2124106, "postDate": "2023-01-31T20:40:52.293Z", "content": "

@carnozhao
\nI'm very surprised you responded to this. Something deeply inappropriate about the question.

", "rawMarkdown": "@carnozhao\nI'm very surprised you responded to this. Something deeply inappropriate about the question." }, { "id": 2124116, "postDate": "2023-01-31T20:44:49.930Z", "content": "

@nikhilmishradev

\n

I am a very confused by the \"suddenly at the end\" concerns.

\n

For many reasons this should be expected.

\n

Blending during merging can have dramatic results if solutions are diverse enough, there's naturally going to be more intense focus at the end for many very obvious reasons, and it's actually very competitive to wait until the end to publish your best models.

\n

I would be shocked if there wasn't sudden jumps near the end of a comp.

", "rawMarkdown": "@nikhilmishradev\n\nI am a very confused by the \"suddenly at the end\" concerns. \n\nFor many reasons this should be expected. \n\nBlending during merging can have dramatic results if solutions are diverse enough, there's naturally going to be more intense focus at the end for many very obvious reasons, and it's actually very competitive to wait until the end to publish your best models.\n\nI would be shocked if there wasn't sudden jumps near the end of a comp." }, { "id": 2124228, "postDate": "2023-01-31T21:59:28Z", "content": "

This is a bit unfair. Copy pasting boilerplate email is hardly a crime, especially for folks who don't speak english well.

\n

I would argue though that they should be more upfront about their situation. Do they have code, can they share it, can they repro, why haven't they submitted, etc.

", "rawMarkdown": "This is a bit unfair. Copy pasting boilerplate email is hardly a crime, especially for folks who don't speak english well.\n\nI would argue though that they should be more upfront about their situation. Do they have code, can they share it, can they repro, why haven't they submitted, etc.", "votes": 1 } ] } ] }, { "id": 2119701, "postDate": "2023-01-29T03:05:44.207Z", "content": "

This says @poteman and @runningz cheated on KDDCup2017.
\nBut I believe this is just a coincidence and they are not involved in this suspicion. 😄
\nhttps://tianchi.aliyun.com/forum/post/1439

", "rawMarkdown": "This says @poteman and @runningz cheated on KDDCup2017.\nBut I believe this is just a coincidence and they are not involved in this suspicion. 😄\nhttps://tianchi.aliyun.com/forum/post/1439", "votes": 14, "replies": [ { "id": 2119706, "postDate": "2023-01-29T03:13:07.497Z", "content": "

what evidence do you have that that the poteman and runningz mentioned in your link, is the same as our poteman and RunningZ ? 😉

", "rawMarkdown": "what evidence do you have that that the poteman and runningz mentioned in your link, is the same as our poteman and RunningZ ? 😉", "votes": 3, "replies": [ { "id": 2119709, "postDate": "2023-01-29T03:19:24.093Z", "content": "

Excactly, this is just a coincidence

\n
3rd place\nHengxing Cai, Sun Yat-sen University, Guangdong Key Laboratory of Intelligent Transportation Systems (Team Leader)\n
\n

https://www.kdd.org/kdd2017/announcements/view/announcing-kdd-cup-2017-highway-tollgates-traffic-flow-prediction

\n
Hi there, I'm Hengxing Cai/蔡恒兴 👋\n
\n

https://github.com/yanqiangmiffy/poteman

", "rawMarkdown": "Excactly, this is just a coincidence\n```\n3rd place\nHengxing Cai, Sun Yat-sen University, Guangdong Key Laboratory of Intelligent Transportation Systems (Team Leader)\n```\nhttps://www.kdd.org/kdd2017/announcements/view/announcing-kdd-cup-2017-highway-tollgates-traffic-flow-prediction\n\n```\nHi there, I'm Hengxing Cai/蔡恒兴 👋\n```\nhttps://github.com/yanqiangmiffy/poteman", "votes": 3 }, { "id": 2119720, "postDate": "2023-01-29T03:36:35.710Z", "content": "

This is just a coincidence
\n
\nhttps://tianchi.aliyun.com/forum/post/2072

", "rawMarkdown": "This is just a coincidence\n\nhttps://tianchi.aliyun.com/forum/post/2072", "votes": 3 }, { "id": 2119722, "postDate": "2023-01-29T03:37:17.657Z", "content": "

RunningZ's email address contained a Chinese name. The third place team has a team member with that exact same name. Yes, it could be a coincidence…

", "rawMarkdown": "RunningZ's email address contained a Chinese name. The third place team has a team member with that exact same name. Yes, it could be a coincidence...", "votes": 6 }, { "id": 2119723, "postDate": "2023-01-29T03:40:01.103Z", "content": "

Yes, that is not the evidence. Coincidence.

", "rawMarkdown": "Yes, that is not the evidence. Coincidence.", "votes": 1 } ] } ] }, { "id": 2119913, "postDate": "2023-01-29T07:54:10.277Z", "content": "

If the credibilty of kaggle medals and titles become questionable, it hurts all kagglers' benifits, expecially for GMs & masters.

\n

If the whole market does not take kaggle achivements as one of the criteriors regards one's ML skills anymore, it is a massive hurt for the people who has been actively working hard on kaggle for YEARS!

", "rawMarkdown": "If the credibilty of kaggle medals and titles become questionable, it hurts all kagglers' benifits, expecially for GMs & masters.\n\nIf the whole market does not take kaggle achivements as one of the criteriors regards one's ML skills anymore, it is a massive hurt for the people who has been actively working hard on kaggle for YEARS!", "votes": 7, "replies": [ { "id": 2120242, "postDate": "2023-01-29T12:59:05.200Z", "content": "

I can't agree more! If this keeps happening, people will stop playing Kaggle. Then all of us lose a fantastic place to learn

", "rawMarkdown": "I can't agree more! If this keeps happening, people will stop playing Kaggle. Then all of us lose a fantastic place to learn", "votes": 1 } ] }, { "id": 2115978, "postDate": "2023-01-26T05:24:09.787Z", "content": "

yes I totally understand and my thoughts are with your team. It feels really bad to know after days of hard work, you come to know someone joining your team might not be playing it fair. Also appreciate the honesty to bring out this, even though it affects your team in this competition.

\n

A problem though I see from other competitors persepective, is as long as that person is in your team, they can download the solution from the submissions tab and share with others, which affects other participants not involved.

", "rawMarkdown": "yes I totally understand and my thoughts are with your team. It feels really bad to know after days of hard work, you come to know someone joining your team might not be playing it fair. Also appreciate the honesty to bring out this, even though it affects your team in this competition.\n\nA problem though I see from other competitors persepective, is as long as that person is in your team, they can download the solution from the submissions tab and share with others, which affects other participants not involved.", "votes": 7, "replies": [ { "id": 2115981, "postDate": "2023-01-26T05:26:48Z", "content": "

Yes, that is very true. We cannot refute this. It is frustrating that we have no means to stop it.

", "rawMarkdown": "Yes, that is very true. We cannot refute this. It is frustrating that we have no means to stop it.", "votes": 4 } ] }, { "id": 2118120, "postDate": "2023-01-27T19:30:03.227Z", "content": "

OMG, he emailed me in the last FB3 competition. I should team up with him, then we can enjoy the shake down together:) From public 2nd to private 86th😹

", "rawMarkdown": "OMG, he emailed me in the last FB3 competition. I should team up with him, then we can enjoy the shake down together:) From public 2nd to private 86th😹", "votes": 5 }, { "id": 2117622, "postDate": "2023-01-27T11:52:12.650Z", "content": "

I teamed up with him in the past PANDA competition.

\n

At the time @kyoshioka47 and I didn't think much of it, but now that I think about it, this is strange.

\n

\"\"

\n

Because if he only wants to do team merge with other teams, he can just tell the leader the team name. There is no need for him to be the team leader.
\nSo why did he want to be the team leader?

\n

Our public LB rank was not high at the time. If our public LB rank was high, would he have wanted to add someone to our team without consulting me and @kyoshioka47 ? Or did he want to remove a sub from the final sub check that was inconvenient for him?

", "rawMarkdown": "I teamed up with him in the past PANDA competition.\n\nAt the time @kyoshioka47 and I didn't think much of it, but now that I think about it, this is strange.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1670024%2F77daa2c8b4d5c28f7b5b2b67f0b770f4%2F2023-01-27%2020.32.01.png?generation=1674819161232660&alt=media)\n\nBecause if he only wants to do team merge with other teams, he can just tell the leader the team name. There is no need for him to be the team leader.\nSo why did he want to be the team leader?\n\nOur public LB rank was not high at the time. If our public LB rank was high, would he have wanted to add someone to our team **without consulting me and @kyoshioka47** ? Or did he want to remove a sub from the final sub check that was inconvenient for him?", "votes": 5, "replies": [ { "id": 2119608, "postDate": "2023-01-28T23:58:41.027Z", "content": "

Is there any evidence to support this?

", "rawMarkdown": "Is there any evidence to support this?", "votes": -19, "replies": [ { "id": 2138059, "postDate": "2023-02-10T14:19:19.233Z", "content": "

\"\"

", "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4645853%2F60cd1ba6f14c5fd91ecc697910b24be9%2Fpoteman.png?generation=1676038757011419&alt=media)", "votes": 2 } ] } ] }, { "id": 2116610, "postDate": "2023-01-26T15:48:13.630Z", "content": "

It's a pity for @mhyodo and the other teammates who worked so hard at this competition. I was also a victim of the same problem as you in the G2Net Gravitational Wave Detection contest and then our private team results were removed (this was expected). Wishing the best for everyone who has worked hard on your team and Kaggle will make decisions that are fairest to all teams on this competition !!!

", "rawMarkdown": "It's a pity for @mhyodo and the other teammates who worked so hard at this competition. I was also a victim of the same problem as you in the G2Net Gravitational Wave Detection contest and then our private team results were removed (this was expected). Wishing the best for everyone who has worked hard on your team and Kaggle will make decisions that are fairest to all teams on this competition !!!", "votes": 5, "replies": [ { "id": 2117080, "postDate": "2023-01-27T00:48:47.267Z", "content": "

Thank you. We have had conversations within the team about \"being prepared to be banned\", and we will follow kaggle team's decision. It was a high tuition fee, hahaha.
\nHowever, we can prove the result of the four of us, (including the code), So we will continue to submit until end of this competition.

", "rawMarkdown": "Thank you. We have had conversations within the team about \"being prepared to be banned\", and we will follow kaggle team's decision. It was a high tuition fee, hahaha. \nHowever, we can prove the result of the four of us, (including the code), So we will continue to submit until end of this competition.", "votes": 5 } ] }, { "id": 2115974, "postDate": "2023-01-26T05:19:31.170Z", "content": "

Thank you for raising this issue. Is this related to this?
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318

", "rawMarkdown": "Thank you for raising this issue. Is this related to this?\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318", "votes": 1, "replies": [ { "id": 2115976, "postDate": "2023-01-26T05:21:54.940Z", "content": "

Yes, I believe this is closely related.

", "rawMarkdown": "Yes, I believe this is closely related.", "votes": 2 } ] }, { "id": 2127811, "postDate": "2023-02-03T07:43:54.430Z", "content": "

just pin this

", "rawMarkdown": "just pin this", "votes": 2 }, { "id": 2123998, "postDate": "2023-01-31T19:36:20.900Z", "content": "

Kudos for not giving up. Perseverance in the face of adversity and such ambiguity is a very admirable trait, especially adversity and ambiguity you brought upon yourselves by being transparent.

\n

I am really pulling for you folks that Kaggle does the right thing here. It would be a signal to folks who come later than being honest and open will be rewarded.

", "rawMarkdown": "Kudos for not giving up. Perseverance in the face of adversity and such ambiguity is a very admirable trait, especially adversity and ambiguity you brought upon yourselves by being transparent.\n\nI am really pulling for you folks that Kaggle does the right thing here. It would be a signal to folks who come later than being honest and open will be rewarded.", "votes": 2 }, { "id": 2119743, "postDate": "2023-01-29T04:29:32.900Z", "content": "

Believe that good will triumph over evil

", "rawMarkdown": "Believe that good will triumph over evil", "votes": 2 }, { "id": 2127821, "postDate": "2023-02-03T08:08:21.090Z", "content": "

Kaggle needs a credit system (Although I will be marked as a dishonest judgment debtors at that time🙈

", "rawMarkdown": "Kaggle needs a credit system (Although I will be marked as a dishonest judgment debtors at that time🙈", "votes": -2, "replies": [ { "id": 2134968, "postDate": "2023-02-08T11:13:55.433Z", "content": "

I am still in favour of your account being deleted for cheating in the toxic comments competition. Even after 5 years I remember you as a cheater who not only was dishonest but also dragged down his whole team! Same will happen to @poteman and @runningz , even after years people will associate your names with cheating.

", "rawMarkdown": "I am still in favour of your account being deleted for cheating in the toxic comments competition. Even after 5 years I remember you as a cheater who not only was dishonest but also dragged down his whole team! Same will happen to @poteman and @runningz , even after years people will associate your names with cheating.", "votes": 7, "replies": [ { "id": 2135361, "postDate": "2023-02-08T15:30:40.127Z", "content": "

I got deleted in this competition without cheating… I have no idea what happened. Maybe I shouldn't team up…

", "rawMarkdown": "I got deleted in this competition without cheating... I have no idea what happened. Maybe I shouldn't team up...", "votes": 3 }, { "id": 2135374, "postDate": "2023-02-08T15:42:17.463Z", "content": "

I feel sorry for your time and energy dedicated on this competition. There is always risk of getting removed when you are merging with people you don't know.

", "rawMarkdown": "I feel sorry for your time and energy dedicated on this competition. There is always risk of getting removed when you are merging with people you don't know.", "votes": 1 }, { "id": 2135827, "postDate": "2023-02-08T22:08:47.293Z", "content": "

Thanks for your understanding. I will continue working on other Kaggle competitions

", "rawMarkdown": "Thanks for your understanding. I will continue working on other Kaggle competitions", "votes": 1 }, { "id": 2136170, "postDate": "2023-02-09T06:21:46.473Z", "content": "

If they really cheat, I will be very upset… That will make me feel that not only can't win the competition, Even cheating methods are not as clever as others

", "rawMarkdown": "If they really cheat, I will be very upset... That will make me feel that not only can't win the competition, Even cheating methods are not as clever as others" } ] } ] }, { "id": 2119578, "postDate": "2023-01-28T22:51:11.503Z", "content": "

It's all speculation without evidence.
\nwhat evidence exists to prove that it's not participating on my own?

", "rawMarkdown": "It's all speculation without evidence. \nwhat evidence exists to prove that it's not participating on my own? ", "votes": -57, "replies": [ { "id": 2119620, "postDate": "2023-01-29T00:05:24.270Z", "content": "

Why did you delete your name from your github? 😄

", "rawMarkdown": "Why did you delete your name from your github? 😄", "votes": 18, "replies": [ { "id": 2119672, "postDate": "2023-01-29T01:42:57.227Z", "content": "

@poteman You can't find good excuse so far? Ok, I'll wait. You can write your excuse here when you come up with.

", "rawMarkdown": "@poteman You can't find good excuse so far? Ok, I'll wait. You can write your excuse here when you come up with.", "votes": 3 } ] }, { "id": 2119637, "postDate": "2023-01-29T00:44:16.913Z", "content": "

It seems to me that it is you who needs to provide the evidence, not us. Under the current circumstances, your company will have no choice but to fire you. You are a great kaggler now ranked in top20. I hope you can provide evidence and avoid getting fired from your company.

", "rawMarkdown": "It seems to me that it is you who needs to provide the evidence, not us. Under the current circumstances, your company will have no choice but to fire you. You are a great kaggler now ranked in top20. I hope you can provide evidence and avoid getting fired from your company.", "votes": 9, "replies": [ { "id": 2119652, "postDate": "2023-01-29T01:15:12.457Z", "content": "

When GM cheated before, his employer, H2O, fired him. And there is too much circumstantial evidence. Even without hard evidence, if 4Paradigm (第四范式) continues to hire you in this situation, 4Paradigm's reputation will suffer greatly. I am sure everyone is watching to see what 4Paradigm will do if you do not provide evidence. You have a solo gold medal. It should be easy for you to present your evidence.

", "rawMarkdown": "When GM cheated before, his employer, H2O, fired him. And there is too much circumstantial evidence. Even without hard evidence, if 4Paradigm (第四范式) continues to hire you in this situation, 4Paradigm's reputation will suffer greatly. I am sure everyone is watching to see what 4Paradigm will do if you do not provide evidence. You have a solo gold medal. It should be easy for you to present your evidence.", "votes": 4 } ] }, { "id": 2119645, "postDate": "2023-01-29T01:07:10.800Z", "content": "

Some people are setting me up because I complained about them with kaggle in previous H&M competitions. They saw the screenshot of my previous email a few days ago.
\n\"\"

", "rawMarkdown": "Some people are setting me up because I complained about them with kaggle in previous H&M competitions. They saw the screenshot of my previous email a few days ago.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F459771%2Fa4e7bd56c20354ffad596f7dc73487ce%2FWechatIMG30438.png?generation=1674954416851321&alt=media)", "votes": -44, "replies": [ { "id": 2119657, "postDate": "2023-01-29T01:22:21.030Z", "content": "

@poteman Hi there, I'm Kazuki Onodera 👋
\nI can reproduce all submissions 😉

", "rawMarkdown": "@poteman Hi there, I'm Kazuki Onodera 👋\nI can reproduce all submissions 😉", "votes": 34 }, { "id": 2124224, "postDate": "2023-01-31T21:55:28.170Z", "content": "

@poteman

\n

Given the array of evidence that has been provided, I'd encourage you provide some kind of defense to the community. I'd also encourage folks to keep an open mind of innocent until proven guilty.

\n

One possible explanation is simply that you've been disorganized and not versioned your code carefully. Honestly, I fall victim to this all the time.

\n

However, I certainly wouldn't team up unless I could openly share my code and repro my results and nor should anyone, or at least they should warn folks of their situation beforehand.

", "rawMarkdown": "@poteman \n\nGiven the array of evidence that has been provided, I'd encourage you provide some kind of defense to the community. I'd also encourage folks to keep an open mind of innocent until proven guilty.\n\nOne possible explanation is simply that you've been disorganized and not versioned your code carefully. Honestly, I fall victim to this **all the time.** \n\nHowever, I certainly wouldn't team up unless I could openly share my code and repro my results and nor should anyone, or at least they should warn folks of their situation beforehand.\n\n\n\n " } ] } ] }, { "id": 2138389, "postDate": "2023-02-10T18:25:59.363Z", "content": "

Update. I've just found the answers here: https://www.kaggle.com/competitions/otto-recommender-system/discussion/381527 \"How cheaters can steal tiers and reputations?\"

\n
\n

I am new to competitions. Please explain to me why sharing code within a team makes a difference compared to sharing predictions. I see that the inability to share code was the cornerstone in this story.
\nMy current understanding is as follows:

\n\n

Is this correct? Am I missing anything? Thanks.

", "rawMarkdown": "Update. I've just found the answers here: https://www.kaggle.com/competitions/otto-recommender-system/discussion/381527 **\"How cheaters can steal tiers and reputations?\"**\n\n------\nI am new to competitions. Please explain to me why sharing code within a team makes a difference compared to sharing predictions. I see that the inability to share code was the cornerstone in this story.\nMy current understanding is as follows:\n- A winning team is required by the rules to submit the code that generated the winning submission, otherwise the team is disqualified. Sharing code within a team gives every team member confidence that their team can reproduce submissions. Thereby ensuring compliance with this particular part of the rules.\n- Sharing code within a team does _not_ guarantee at all that no cheating takes place. It is possible that an owner of multiple accounts takes code shared with them in one team and passes it on as their own to another team. Moreover, using many fake accounts the cheater can climb the ladder by joining successively stronger teams and end up in the top without even having compute power to execute the code they plagiarised. If the stolen code is not used by another winning team, they may even get away with their cheating.\n\nIs this correct? Am I missing anything? Thanks." }, { "id": 2123339, "postDate": "2023-01-31T12:58:15.223Z", "content": "

As a new-in-town in community, i was shocked that such unfair issues are raging and destroying this nice community. And you actually remind me of some very suspicious topics in \"XIAO HONG SHU\" saying that they guarantee sliver/bronze if you pay them….

", "rawMarkdown": "As a new-in-town in community, i was shocked that such unfair issues are raging and destroying this nice community. And you actually remind me of some very suspicious topics in \"XIAO HONG SHU\" saying that they guarantee sliver/bronze if you pay them...." }, { "id": 2123194, "postDate": "2023-01-31T11:23:23.083Z", "content": "

It doesn't feel good to see kaggle's platform infiltrated by unfair mean

", "rawMarkdown": "It doesn't feel good to see kaggle's platform infiltrated by unfair mean" }, { "id": 2118744, "postDate": "2023-01-28T08:36:53.377Z", "rawMarkdown": "", "votes": 1, "isDeleted": true, "replies": [ { "id": 2118935, "postDate": "2023-01-28T11:53:35.707Z", "rawMarkdown": "", "votes": 1, "replies": [ { "id": 2118963, "postDate": "2023-01-28T12:18:08.253Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2118966, "postDate": "2023-01-28T12:20:47.583Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2118973, "postDate": "2023-01-28T12:25:44.860Z", "rawMarkdown": "", "votes": -1, "isDeleted": true }, { "id": 2118979, "postDate": "2023-01-28T12:27:21.887Z", "rawMarkdown": "", "votes": -1, "isDeleted": true }, { "id": 2119053, "postDate": "2023-01-28T13:17:51.020Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ] } ] } ], "comments": [ { "id": 2119716, "author_name": "narsil (jobs-in-data.com)", "author_url": "", "post_date": "2023-01-29T03:33:38.697000", "content": "

The sad part is that getting a gold medal in a Kaggle competition is very hard. Getting a solo gold is extremely hard. How many honest people were denied gold medals in the past because of cheating? People who spent countless hours and nights training their models…

\n

The very existence of such schemes destroys the fun and competitive aspect of Kaggle competitions. Kaggle should treat those matters extremely seriously if they want to avoid the dilution of the Kaggle Competitions brand, especially the competitions Grandmaster brand, which now stands for quite a lot (there are companies recruiting based on obtaining the GM title).

", "votes": 35, "replies": [ { "id": 2120307, "author_name": "Anthony Chiu", "author_url": "", "post_date": "2023-01-29T13:59:44.407000", "content": "

Thanks for your words. Being a GM is on my before-30 bucket list but sadly I got 2 silvers last year:(. I still have 9 months to achieve this haha, I think I will stick with code competition…

", "votes": 4, "replies": [ { "id": 2120960, "author_name": "narsil (jobs-in-data.com)", "author_url": "", "post_date": "2023-01-29T23:21:51.653000", "content": "

Keeping my fingers crossed for you. I still remember your starting at the HomeCredit competition.

", "votes": 1, "replies": [] } ] } ] }, { "id": 2117522, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-27T10:32:31.633000", "content": "

I'm just wondering why he deleted his name from his github yesterday.
\n
\nhttps://github.com/poteman/poteman/commit/de214efb1495151385fc199dd8d885541fabdf9e

", "votes": 31, "replies": [ { "id": 2117716, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-27T13:41:43.060000", "content": "

@poteman Now we can't see your name. Well done😄👍
\n

", "votes": 15, "replies": [ { "id": 2118433, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-28T02:27:53.297000", "content": "

@poteman We can still develop your \"Hi there\" OSS. This is the beauty of OSS.
\n
\nhttps://github.com/yanqiangmiffy/poteman

", "votes": 19, "replies": [] } ] } ] }, { "id": 2116033, "author_name": "arutema47", "author_url": "", "post_date": "2023-01-26T06:24:31.360000", "content": "

I teamed up with him in the past PANDA competition.
\nHe said he could not share his training code due to a paper submission, and we thought that was ok since our private 1st sub did not contain his model. So we left out his codes in the summary and the winner's call.

\n

Looking at this incident, I looked back to his notebook and found a strange match between his code and the Public 3rd, Private 20th team's solution. (https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/169437)

\n

This model (MultiheadAttentionClassifier) is quite unique to my knowledge, and I could not find other code match after searching in the github. Only SHUJUN's codes appear after the search, and it wasn't public during the competition was held. (There might be more than 2 suspects?)

\n

\"\"

", "votes": 29, "replies": [ { "id": 2116044, "author_name": "arutema47", "author_url": "", "post_date": "2023-01-26T06:37:15.517000", "content": "

The network part is quite identical as well..

\n

https://github.com/Shujun-He/PANDA/blob/e0327b1a604b4c1444827262c931daa47c10855c/pandalayer1test8_s64/Network.py#L83

", "votes": 2, "replies": [] }, { "id": 2116046, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2023-01-26T06:40:20.427000", "content": "

Or he didn't even have code at that time. After the competition ended, he copied someone elses solution in order to drop suspicions.

", "votes": 3, "replies": [ { "id": 2116047, "author_name": "arutema47", "author_url": "", "post_date": "2023-01-26T06:42:09.640000", "content": "

No, this was a notebook competition and the above are his codes* during* the competition was held.

", "votes": 2, "replies": [] } ] }, { "id": 2118580, "author_name": "Shujun", "author_url": "", "post_date": "2023-01-28T06:02:24.977000", "content": "

Thanks for bringing this to light @kyoshioka47 . It is indeed very strange that @poteman had the exact same code as my solution in PANDA. In fact, my account was (wrongly) banned as a result of a major compliance action following SIIM-ISIC Melanoma Classification, because Kaggle thought I sold my code to a member of angtk. You can check out @group16 's excellent investigative work about angtk here: https://www.kaggle.com/competitions/siim-isic-melanoma-classification/discussion/175156.

\n

I knew for a fact that I did not have anything to do with angtk, so I immediately appealed to Kaggle to get my account restored. Keep in mind that my real identity on Linkedin is linked to my Kaggle account so you can see how serious that was for me: effectively I was labeled a cheater and I needed to clear my name. Luckily, after a few email exchanges, I was able to convince Kaggle to restore my account shortly after. I think Kaggle did not have clear evidence and only linked me with angtk since I released my code on github.

\n

After speaking to @group16 , we think that @ruozha001 was connected with @poteman and sold my code to @poteman after she got her hands on it. You can also check out the post made by @group16 (https://www.kaggle.com/competitions/otto-recommender-system/discussion/381588).

", "votes": 21, "replies": [ { "id": 2119123, "author_name": "CroDoc", "author_url": "", "post_date": "2023-01-28T14:01:01.487000", "content": "

I am afraid to merge with anyone who I don't know now 🤯

", "votes": 0, "replies": [] } ] }, { "id": 2119573, "author_name": "", "author_url": "", "post_date": "2023-01-28T22:41:34.703000", "content": "

This competition was held 3 years ago and I have forgotten many details.

\n
    \n
  1. What evidence is there that this part of the code belongs to me? Especially I already have the codes before the game ends.
    2. \n
  2. If the first point is true, what evidence is there to prove that the codes are not written by myself?
    3. \n
", "votes": -10, "replies": [ { "id": 2119632, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T00:23:56.587000", "content": "

@poteman Hi there, I'm Kazuki Onodera👋

\n
\n

I already have the codes before the game ends.

\n
\n

So you're saying this is a coincidence. I think this is bad move. You should have said \"I copied his code after this comp\".

", "votes": 0, "replies": [] }, { "id": 2119668, "author_name": "@kaggleqrdl", "author_url": "", "post_date": "2023-01-29T01:40:23.223000", "content": "

He may be saying : \"What evidence is there .. that I already have the codes before the games ends?\"

\n

I think folks are getting silly, maybe take a beat and consider what are reasonable assumptions and what might be a bit unfair.

\n

Tracking the provenance of code is not a simple task that can be done off the cuff. It requires an immense amount of effort. I've found things on git only after a great deal of digging, it's very deep web and not something that will necessarily show up on a one off google search.

\n

Things also get deleted but still stay in personal repos. There can be all sorts of explanations.

", "votes": 0, "replies": [] }, { "id": 2119763, "author_name": "arutema47", "author_url": "", "post_date": "2023-01-29T04:47:19.590000", "content": "

Well, his codes are submitted during the completion and the Kaggle team can easily check this. If the team can mail me about this, I am happy to point out.

", "votes": 1, "replies": [] }, { "id": 2119802, "author_name": "Gilles Vandewiele", "author_url": "", "post_date": "2023-01-29T05:46:43.893000", "content": "
\n

This competition was held 3 years ago and I have forgotten many details.

\n
    \n
  1. What evidence is there that this part of the code belongs to me? Especially I already have the codes before the game ends.
    2. \n
  2. If the first point is true, what evidence is there to prove that the codes are not written by myself?
    3. \n
\n
\n
    \n
  1. It's on your github… 🤦‍♂️🍩
    2. \n
  2. While not being caught red-handed, the probability of sharing so much similarities (it's the exact same) between two pieces of code is astronomically small, except when this is taken from a public resource (but this is not the case). Both codes even contain no spaces before & after the =. Your public notebooks on Kaggle, however, seem to always include spaces?
    3. \n
", "votes": 4, "replies": [] } ] }, { "id": 2119833, "author_name": "fam_taro", "author_url": "", "post_date": "2023-01-29T06:18:57.313000", "content": "

Oh, for some reason you seem to have deleted all your submission notebooks for panda competition.

\n

\"\"

\n

This will reduce the evidence to prove your innocence.

\n

But don't worry. The code is still on the kaggle server because it has been submitted. Let's check it with kaggle staff.
\nAt the same time, comparing it to @shujun717 's subs will prove your innocence. Good for you👍

", "votes": 12, "replies": [] }, { "id": 2122932, "author_name": "Gleb", "author_url": "", "post_date": "2023-01-31T08:24:33.340000", "content": "

@kyoshioka47 @yukkyo @shujun717 I think i got it, what happend in PANDA. For some time i cant wrap my head around why there are fake teams above poteman on public:

\n
    \n
  1. poteman joins with K and Y
    2. \n
  2. ruozha001 joins S. Clearly its a fake \"female\" account to target teams that can do well and leak results afterwards
    3. \n
  3. S goes high on public, ruozha001 leaks exact ensemble to other \"boost\" teams, they are fake-users only
    4. \n
  4. \"boost\" teams try to worse the S score and post it, you can still see some of those teams on LB, not banned
    5. \n
  5. poteman at the same time is stuck with K and Y, they are 20-ish on public. He understands that stealing whole S ensemble is super-risky, as he has K&Y in his team, they are not fake
    6. \n
  6. poteman steals just LB best single model from S ensemble, its around .910, thats all he can do, he is upset at 20th place with those \"losers\" K&Y
    7. \n
  7. K&Y just win competition without poteman stolen model.
    8. \n
\n

If all of that true, then S should have best single model around .910 LB, bb being resnext50-whatever it was.

", "votes": 5, "replies": [] } ] }, { "id": 2116011, "author_name": "YYama", "author_url": "", "post_date": "2023-01-26T06:08:31.477000", "content": "

In another competition, I have received a team merge request from two of alleged kagglers at the almost same time. I was in the gold zone at the time and got an email from the master first.

\n

He had never submitted at the competition but had a great record in the past competitions and we were considering merging, but we decided not to because we considered it a risk to invite a stranger who had never submitted.
\nShortly thereafter, we received a merge request from a grandmaster. Surprisingly, with only a few exceptions, the email was almost identical to the email from the master in terms of the position of the surprise mark and the wording of the text.

\n

I suspected that there might be some connection between the two kagglers. Then I discovered that they had both performed well together quite often, even though they had been on different teams in past competitions. I suspected some kind of private sharing or even the same person.

\n

I did not report this about it as it was only a circumstantial suspicion. However now, I strongly suspect that these two are in cahoots, or are the same person, and are acting to subvert the other team members to their own advantage.

\n

I have teamed up with makotu in the past. He was a very enthusiastic kaggler, quick to share good experimental code, and extremely talented. I posted this in the hope that my experience could reinforce makotu's assertion.

", "votes": 17, "replies": [ { "id": 2116019, "author_name": "YYama", "author_url": "", "post_date": "2023-01-26T06:16:00.700000", "content": "

This mail is from the grand master

\n
\n

Hi, YYama.

\n

Hope you are doing well!
\n I am writing this to you to check if you are interested in teaming up with me for the competition \"RSNA 2022 Cervical Spine Fracture Detection\".
\n We can make a team together and get better score.

\n

Looking forward to your reply.

\n
\n

This mail is from the master

\n
\n

Hi, YYama
\n Hope you are doing well!
\n I am writing this to you to check if you are interested in teaming up with me for the competition RSNA

\n

About device:
\n 16 * A100 gpu and colob pro, and I can also rent gpu if needed.

\n

We can make a team together and get better score.
\n Looking forward to your reply.

\n
\n

They wouldn't match so much by coincidence. But if I had only received an email from one of them, I would not have suspected cheating.

", "votes": 6, "replies": [ { "id": 2116024, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-26T06:18:36.063000", "content": "

Even I got the mails in close proximity for this competition. I am sure, if I teamed up with one, the other would not have mailed

", "votes": 2, "replies": [] }, { "id": 2116048, "author_name": "zakopuro", "author_url": "", "post_date": "2023-01-26T06:42:20.247000", "content": "

I received a similar email.

\n
Hi, zakopuro.\n\nI hope all is well!\nI'm writing this to you to check if you are interested in teaming up with me for the competition 'American Express'. \nWe can make a team together and get better score.\n\nLooking forward to your reply.\n
\n

When I teamed up with him in the past, he said his PC was broken and he couldn't share code.
\nAgain, was he unable to share the code because his PC was broken?😂

", "votes": 8, "replies": [] }, { "id": 2116163, "author_name": "Carno Zhao", "author_url": "", "post_date": "2023-01-26T09:07:10.960000", "content": "

Same here.

\n

In HPA+HUBMAP last year, our team also recieved the similar team-merge email from him. At that time, we were in gold zone and he did not even make a single submission!

\n

\"\"

", "votes": 6, "replies": [] }, { "id": 2116249, "author_name": "HAO", "author_url": "", "post_date": "2023-01-26T10:53:40.647000", "content": "

I received a very similar Email from him in H&M competition two weeks before the submission deadline, when I was rankinng 3rd and he didn't even submit anything. \"\"

", "votes": 2, "replies": [] }, { "id": 2116330, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-26T12:27:03.883000", "content": "

@carnozhao Just curious, can your team reproduce the best submission from scratch?

", "votes": 3, "replies": [] }, { "id": 2116337, "author_name": "Carno Zhao", "author_url": "", "post_date": "2023-01-26T12:32:00.290000", "content": "

Yes, if you ignore model training randomness

", "votes": 0, "replies": [] }, { "id": 2116347, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-26T12:39:35.107000", "content": "

That means you already checked reproducibility of your teammate codes?

", "votes": 0, "replies": [] }, { "id": 2116352, "author_name": "Carno Zhao", "author_url": "", "post_date": "2023-01-26T12:41:58.740000", "content": "

yes, most of code is written by me

", "votes": 2, "replies": [] }, { "id": 2116356, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-26T12:45:59.913000", "content": "

Thanks, I'm relieved to hear that. In this situation, we can't trust random guys's submission file.

", "votes": 1, "replies": [] }, { "id": 2116362, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-26T12:50:49.047000", "content": "

what I am concerned is about, is it possible some people share the team's solution outside it ? I don't intend to blame anyone, but looks like this could be the case with these two coming up, so suddenly at the end of the competition. I think their first few submissions needs to be dug up, is it possible to identify the source. Someone within a high scoring team, probably shared their submisison with these 2 guys

", "votes": 3, "replies": [] }, { "id": 2116368, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2023-01-26T12:55:24.507000", "content": "

That's what I thought. We still don't know how they can always join the leaderboard in gold zone. Someone from top teams might be leaking their submissions.

", "votes": 3, "replies": [] }, { "id": 2124106, "author_name": "@kaggleqrdl", "author_url": "", "post_date": "2023-01-31T20:40:52.293000", "content": "

@carnozhao
\nI'm very surprised you responded to this. Something deeply inappropriate about the question.

", "votes": 0, "replies": [] }, { "id": 2124116, "author_name": "@kaggleqrdl", "author_url": "", "post_date": "2023-01-31T20:44:49.930000", "content": "

@nikhilmishradev

\n

I am a very confused by the \"suddenly at the end\" concerns.

\n

For many reasons this should be expected.

\n

Blending during merging can have dramatic results if solutions are diverse enough, there's naturally going to be more intense focus at the end for many very obvious reasons, and it's actually very competitive to wait until the end to publish your best models.

\n

I would be shocked if there wasn't sudden jumps near the end of a comp.

", "votes": 0, "replies": [] }, { "id": 2124228, "author_name": "@kaggleqrdl", "author_url": "", "post_date": "2023-01-31T21:59:28", "content": "

This is a bit unfair. Copy pasting boilerplate email is hardly a crime, especially for folks who don't speak english well.

\n

I would argue though that they should be more upfront about their situation. Do they have code, can they share it, can they repro, why haven't they submitted, etc.

", "votes": 1, "replies": [] } ] } ] }, { "id": 2119701, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T03:05:44.207000", "content": "

This says @poteman and @runningz cheated on KDDCup2017.
\nBut I believe this is just a coincidence and they are not involved in this suspicion. 😄
\nhttps://tianchi.aliyun.com/forum/post/1439

", "votes": 14, "replies": [ { "id": 2119706, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-29T03:13:07.497000", "content": "

what evidence do you have that that the poteman and runningz mentioned in your link, is the same as our poteman and RunningZ ? 😉

", "votes": 3, "replies": [ { "id": 2119709, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T03:19:24.093000", "content": "

Excactly, this is just a coincidence

\n
3rd place\nHengxing Cai, Sun Yat-sen University, Guangdong Key Laboratory of Intelligent Transportation Systems (Team Leader)\n
\n

https://www.kdd.org/kdd2017/announcements/view/announcing-kdd-cup-2017-highway-tollgates-traffic-flow-prediction

\n
Hi there, I'm Hengxing Cai/蔡恒兴 👋\n
\n

https://github.com/yanqiangmiffy/poteman

", "votes": 3, "replies": [] }, { "id": 2119720, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T03:36:35.710000", "content": "

This is just a coincidence
\n
\nhttps://tianchi.aliyun.com/forum/post/2072

", "votes": 3, "replies": [] }, { "id": 2119722, "author_name": "YYama", "author_url": "", "post_date": "2023-01-29T03:37:17.657000", "content": "

RunningZ's email address contained a Chinese name. The third place team has a team member with that exact same name. Yes, it could be a coincidence…

", "votes": 6, "replies": [] }, { "id": 2119723, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T03:40:01.103000", "content": "

Yes, that is not the evidence. Coincidence.

", "votes": 1, "replies": [] } ] } ] }, { "id": 2119913, "author_name": "BUUMOO", "author_url": "", "post_date": "2023-01-29T07:54:10.277000", "content": "

If the credibilty of kaggle medals and titles become questionable, it hurts all kagglers' benifits, expecially for GMs & masters.

\n

If the whole market does not take kaggle achivements as one of the criteriors regards one's ML skills anymore, it is a massive hurt for the people who has been actively working hard on kaggle for YEARS!

", "votes": 7, "replies": [ { "id": 2120242, "author_name": "william.wu", "author_url": "", "post_date": "2023-01-29T12:59:05.200000", "content": "

I can't agree more! If this keeps happening, people will stop playing Kaggle. Then all of us lose a fantastic place to learn

", "votes": 1, "replies": [] } ] }, { "id": 2115978, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-26T05:24:09.787000", "content": "

yes I totally understand and my thoughts are with your team. It feels really bad to know after days of hard work, you come to know someone joining your team might not be playing it fair. Also appreciate the honesty to bring out this, even though it affects your team in this competition.

\n

A problem though I see from other competitors persepective, is as long as that person is in your team, they can download the solution from the submissions tab and share with others, which affects other participants not involved.

", "votes": 7, "replies": [ { "id": 2115981, "author_name": "Makotu", "author_url": "", "post_date": "2023-01-26T05:26:48", "content": "

Yes, that is very true. We cannot refute this. It is frustrating that we have no means to stop it.

", "votes": 4, "replies": [] } ] }, { "id": 2118120, "author_name": "william.wu", "author_url": "", "post_date": "2023-01-27T19:30:03.227000", "content": "

OMG, he emailed me in the last FB3 competition. I should team up with him, then we can enjoy the shake down together:) From public 2nd to private 86th😹

", "votes": 5, "replies": [] }, { "id": 2117622, "author_name": "fam_taro", "author_url": "", "post_date": "2023-01-27T11:52:12.650000", "content": "

I teamed up with him in the past PANDA competition.

\n

At the time @kyoshioka47 and I didn't think much of it, but now that I think about it, this is strange.

\n

\"\"

\n

Because if he only wants to do team merge with other teams, he can just tell the leader the team name. There is no need for him to be the team leader.
\nSo why did he want to be the team leader?

\n

Our public LB rank was not high at the time. If our public LB rank was high, would he have wanted to add someone to our team without consulting me and @kyoshioka47 ? Or did he want to remove a sub from the final sub check that was inconvenient for him?

", "votes": 5, "replies": [ { "id": 2119608, "author_name": "", "author_url": "", "post_date": "2023-01-28T23:58:41.027000", "content": "

Is there any evidence to support this?

", "votes": -19, "replies": [ { "id": 2138059, "author_name": "Ali Abdin", "author_url": "", "post_date": "2023-02-10T14:19:19.233000", "content": "

\"\"

", "votes": 2, "replies": [] } ] } ] }, { "id": 2116610, "author_name": "KhanhVD", "author_url": "", "post_date": "2023-01-26T15:48:13.630000", "content": "

It's a pity for @mhyodo and the other teammates who worked so hard at this competition. I was also a victim of the same problem as you in the G2Net Gravitational Wave Detection contest and then our private team results were removed (this was expected). Wishing the best for everyone who has worked hard on your team and Kaggle will make decisions that are fairest to all teams on this competition !!!

", "votes": 5, "replies": [ { "id": 2117080, "author_name": "Makotu", "author_url": "", "post_date": "2023-01-27T00:48:47.267000", "content": "

Thank you. We have had conversations within the team about \"being prepared to be banned\", and we will follow kaggle team's decision. It was a high tuition fee, hahaha.
\nHowever, we can prove the result of the four of us, (including the code), So we will continue to submit until end of this competition.

", "votes": 5, "replies": [] } ] }, { "id": 2115974, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-26T05:19:31.170000", "content": "

Thank you for raising this issue. Is this related to this?
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318

", "votes": 1, "replies": [ { "id": 2115976, "author_name": "Makotu", "author_url": "", "post_date": "2023-01-26T05:21:54.940000", "content": "

Yes, I believe this is closely related.

", "votes": 2, "replies": [] } ] }, { "id": 2127811, "author_name": "Hikkiiiiiiiii", "author_url": "", "post_date": "2023-02-03T07:43:54.430000", "content": "

just pin this

", "votes": 2, "replies": [] }, { "id": 2123998, "author_name": "@kaggleqrdl", "author_url": "", "post_date": "2023-01-31T19:36:20.900000", "content": "

Kudos for not giving up. Perseverance in the face of adversity and such ambiguity is a very admirable trait, especially adversity and ambiguity you brought upon yourselves by being transparent.

\n

I am really pulling for you folks that Kaggle does the right thing here. It would be a signal to folks who come later than being honest and open will be rewarded.

", "votes": 2, "replies": [] }, { "id": 2119743, "author_name": "Shiner", "author_url": "", "post_date": "2023-01-29T04:29:32.900000", "content": "

Believe that good will triumph over evil

", "votes": 2, "replies": [] }, { "id": 2127821, "author_name": "Hikkiiiiiiiii", "author_url": "", "post_date": "2023-02-03T08:08:21.090000", "content": "

Kaggle needs a credit system (Although I will be marked as a dishonest judgment debtors at that time🙈

", "votes": -2, "replies": [ { "id": 2134968, "author_name": "Dieter", "author_url": "", "post_date": "2023-02-08T11:13:55.433000", "content": "

I am still in favour of your account being deleted for cheating in the toxic comments competition. Even after 5 years I remember you as a cheater who not only was dishonest but also dragged down his whole team! Same will happen to @poteman and @runningz , even after years people will associate your names with cheating.

", "votes": 7, "replies": [ { "id": 2135361, "author_name": "william.wu", "author_url": "", "post_date": "2023-02-08T15:30:40.127000", "content": "

I got deleted in this competition without cheating… I have no idea what happened. Maybe I shouldn't team up…

", "votes": 3, "replies": [] }, { "id": 2135374, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2023-02-08T15:42:17.463000", "content": "

I feel sorry for your time and energy dedicated on this competition. There is always risk of getting removed when you are merging with people you don't know.

", "votes": 1, "replies": [] }, { "id": 2135827, "author_name": "william.wu", "author_url": "", "post_date": "2023-02-08T22:08:47.293000", "content": "

Thanks for your understanding. I will continue working on other Kaggle competitions

", "votes": 1, "replies": [] }, { "id": 2136170, "author_name": "Hikkiiiiiiiii", "author_url": "", "post_date": "2023-02-09T06:21:46.473000", "content": "

If they really cheat, I will be very upset… That will make me feel that not only can't win the competition, Even cheating methods are not as clever as others

", "votes": 0, "replies": [] } ] } ] }, { "id": 2119578, "author_name": "", "author_url": "", "post_date": "2023-01-28T22:51:11.503000", "content": "

It's all speculation without evidence.
\nwhat evidence exists to prove that it's not participating on my own?

", "votes": -57, "replies": [ { "id": 2119620, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T00:05:24.270000", "content": "

Why did you delete your name from your github? 😄

", "votes": 18, "replies": [ { "id": 2119672, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T01:42:57.227000", "content": "

@poteman You can't find good excuse so far? Ok, I'll wait. You can write your excuse here when you come up with.

", "votes": 3, "replies": [] } ] }, { "id": 2119637, "author_name": "YujiAriyasu", "author_url": "", "post_date": "2023-01-29T00:44:16.913000", "content": "

It seems to me that it is you who needs to provide the evidence, not us. Under the current circumstances, your company will have no choice but to fire you. You are a great kaggler now ranked in top20. I hope you can provide evidence and avoid getting fired from your company.

", "votes": 9, "replies": [ { "id": 2119652, "author_name": "YujiAriyasu", "author_url": "", "post_date": "2023-01-29T01:15:12.457000", "content": "

When GM cheated before, his employer, H2O, fired him. And there is too much circumstantial evidence. Even without hard evidence, if 4Paradigm (第四范式) continues to hire you in this situation, 4Paradigm's reputation will suffer greatly. I am sure everyone is watching to see what 4Paradigm will do if you do not provide evidence. You have a solo gold medal. It should be easy for you to present your evidence.

", "votes": 4, "replies": [] } ] }, { "id": 2119645, "author_name": "", "author_url": "", "post_date": "2023-01-29T01:07:10.800000", "content": "

Some people are setting me up because I complained about them with kaggle in previous H&M competitions. They saw the screenshot of my previous email a few days ago.
\n\"\"

", "votes": -44, "replies": [ { "id": 2119657, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-29T01:22:21.030000", "content": "

@poteman Hi there, I'm Kazuki Onodera 👋
\nI can reproduce all submissions 😉

", "votes": 34, "replies": [] }, { "id": 2124224, "author_name": "@kaggleqrdl", "author_url": "", "post_date": "2023-01-31T21:55:28.170000", "content": "

@poteman

\n

Given the array of evidence that has been provided, I'd encourage you provide some kind of defense to the community. I'd also encourage folks to keep an open mind of innocent until proven guilty.

\n

One possible explanation is simply that you've been disorganized and not versioned your code carefully. Honestly, I fall victim to this all the time.

\n

However, I certainly wouldn't team up unless I could openly share my code and repro my results and nor should anyone, or at least they should warn folks of their situation beforehand.

", "votes": 0, "replies": [] } ] } ] }, { "id": 2138389, "author_name": "Andrey", "author_url": "", "post_date": "2023-02-10T18:25:59.363000", "content": "

Update. I've just found the answers here: https://www.kaggle.com/competitions/otto-recommender-system/discussion/381527 \"How cheaters can steal tiers and reputations?\"

\n
\n

I am new to competitions. Please explain to me why sharing code within a team makes a difference compared to sharing predictions. I see that the inability to share code was the cornerstone in this story.
\nMy current understanding is as follows:

\n\n

Is this correct? Am I missing anything? Thanks.

", "votes": 0, "replies": [] }, { "id": 2123339, "author_name": "AnrDaku", "author_url": "", "post_date": "2023-01-31T12:58:15.223000", "content": "

As a new-in-town in community, i was shocked that such unfair issues are raging and destroying this nice community. And you actually remind me of some very suspicious topics in \"XIAO HONG SHU\" saying that they guarantee sliver/bronze if you pay them….

", "votes": 0, "replies": [] }, { "id": 2123194, "author_name": "Jacobvvvv", "author_url": "", "post_date": "2023-01-31T11:23:23.083000", "content": "

It doesn't feel good to see kaggle's platform infiltrated by unfair mean

", "votes": 0, "replies": [] }, { "id": 2118744, "author_name": "", "author_url": "", "post_date": "2023-01-28T08:36:53.377000", "content": "", "votes": 1, "replies": [ { "id": 2118935, "author_name": "", "author_url": "", "post_date": "2023-01-28T11:53:35.707000", "content": "", "votes": 1, "replies": [ { "id": 2118963, "author_name": "", "author_url": "", "post_date": "2023-01-28T12:18:08.253000", "content": "", "votes": 0, "replies": [] }, { "id": 2118966, "author_name": "", "author_url": "", "post_date": "2023-01-28T12:20:47.583000", "content": "", "votes": 0, "replies": [] }, { "id": 2118973, "author_name": "", "author_url": "", "post_date": "2023-01-28T12:25:44.860000", "content": "", "votes": -1, "replies": [] }, { "id": 2118979, "author_name": "", "author_url": "", "post_date": "2023-01-28T12:27:21.887000", "content": "", "votes": -1, "replies": [] }, { "id": 2119053, "author_name": "", "author_url": "", "post_date": "2023-01-28T13:17:51.020000", "content": "", "votes": -1, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2115966": "I know many of you are curious, so let me explain the current status of our team.\n\n・First of all, our team score is a merge of the 4 of us. We can release the results, including the code, after the competition is over.\n\n・Because there was a fear that specific person were not participating on their own, We asked him to share the code they had written as of yesterday morning. He said he achieved 0.596 himself, but did not share that code(for now, We don't know how he ended up with that result.) As a result, he was not able to share it with us, so we decided not be cooperating with that person. Specifically, we(4 of us) move other slack, and we do not have access to each other's information.\n\n・Of course, we do not share the code/data to him.\n\nI hope you understand that this is all we can do at this point, and that at this point we are basically four people thinking and working together.\n\nIt is really sad to see such allegations in a competition for model accuracy. I too think that not only the prize winners, but also those who enter the gold zone should be provided with models and codes that produce the same results.\n\nI leave the decision to kaggle Team. \nWe will be happy to answer any questions you may have.\n\nPS：\nTo prevent this from happening in the future, slack communication with him will be open to public. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F397595%2F4247c019da1b917bd7d18569426fd2c9%2Fslack_1.png?generation=1674712574709325&alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F397595%2F2bd07680a672bfbcff0f156af9a8474c%2Fslack_v2.png?generation=1674712590241405&alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F397595%2F95628b01fdc797181330a9e83d82d1f3%2Fslack_3.png?generation=1674712604829685&alt=media)", "2119716": "The sad part is that getting a gold medal in a Kaggle competition is very hard. Getting a solo gold is extremely hard. How many honest people were denied gold medals in the past because of cheating? People who spent countless hours and nights training their models...\n\nThe very existence of such schemes destroys the fun and competitive aspect of Kaggle competitions. Kaggle should treat those matters extremely seriously if they want to avoid the dilution of the Kaggle Competitions brand, especially the competitions Grandmaster brand, which now stands for quite a lot (there are companies recruiting based on obtaining the GM title).", "2117522": "I'm just wondering why he deleted his name from his github yesterday.\n\nhttps://github.com/poteman/poteman/commit/de214efb1495151385fc199dd8d885541fabdf9e ", "2116033": "I teamed up with him in the past PANDA competition.\nHe said he could not share his training code due to a paper submission, and we thought that was ok since our private 1st sub did not contain his model. So we left out his codes in the summary and the winner's call.\n\nLooking at this incident, I looked back to his notebook and found a strange match between his code and the Public 3rd, Private 20th team's solution. (https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/169437)\n\nThis model (MultiheadAttentionClassifier) is quite unique to my knowledge, and I could not find other code match after searching in the github. Only SHUJUN's codes appear after the search, and it wasn't public during the competition was held. (There might be more than 2 suspects?)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3508221%2F6bc0a5590056ca45b967bb544bf9e571%2F2023-01-26%20152028.png?generation=1674714058896654&alt=media)", "2116011": "In another competition, I have received a team merge request from two of alleged kagglers at the almost same time. I was in the gold zone at the time and got an email from the master first.\n\nHe had never submitted at the competition but had a great record in the past competitions and we were considering merging, but we decided not to because we considered it a risk to invite a stranger who had never submitted.\nShortly thereafter, we received a merge request from a grandmaster. Surprisingly, with only a few exceptions, the email was almost identical to the email from the master in terms of the position of the surprise mark and the wording of the text.\n\nI suspected that there might be some connection between the two kagglers. Then I discovered that they had both performed well together quite often, even though they had been on different teams in past competitions. I suspected some kind of private sharing or even the same person.\n\nI did not report this about it as it was only a circumstantial suspicion. However now, I strongly suspect that these two are in cahoots, or are the same person, and are acting to subvert the other team members to their own advantage.\n\nI have teamed up with makotu in the past. He was a very enthusiastic kaggler, quick to share good experimental code, and extremely talented. I posted this in the hope that my experience could reinforce makotu's assertion.", "2119701": "This says @poteman and @runningz cheated on KDDCup2017.\nBut I believe this is just a coincidence and they are not involved in this suspicion. 😄\nhttps://tianchi.aliyun.com/forum/post/1439", "2119913": "If the credibilty of kaggle medals and titles become questionable, it hurts all kagglers' benifits, expecially for GMs & masters.\n\nIf the whole market does not take kaggle achivements as one of the criteriors regards one's ML skills anymore, it is a massive hurt for the people who has been actively working hard on kaggle for YEARS!", "2115978": "yes I totally understand and my thoughts are with your team. It feels really bad to know after days of hard work, you come to know someone joining your team might not be playing it fair. Also appreciate the honesty to bring out this, even though it affects your team in this competition.\n\nA problem though I see from other competitors persepective, is as long as that person is in your team, they can download the solution from the submissions tab and share with others, which affects other participants not involved.", "2118120": "OMG, he emailed me in the last FB3 competition. I should team up with him, then we can enjoy the shake down together:) From public 2nd to private 86th😹", "2117622": "I teamed up with him in the past PANDA competition.\n\nAt the time @kyoshioka47 and I didn't think much of it, but now that I think about it, this is strange.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1670024%2F77daa2c8b4d5c28f7b5b2b67f0b770f4%2F2023-01-27%2020.32.01.png?generation=1674819161232660&alt=media)\n\nBecause if he only wants to do team merge with other teams, he can just tell the leader the team name. There is no need for him to be the team leader.\nSo why did he want to be the team leader?\n\nOur public LB rank was not high at the time. If our public LB rank was high, would he have wanted to add someone to our team **without consulting me and @kyoshioka47** ? Or did he want to remove a sub from the final sub check that was inconvenient for him?", "2116610": "It's a pity for @mhyodo and the other teammates who worked so hard at this competition. I was also a victim of the same problem as you in the G2Net Gravitational Wave Detection contest and then our private team results were removed (this was expected). Wishing the best for everyone who has worked hard on your team and Kaggle will make decisions that are fairest to all teams on this competition !!!", "2115974": "Thank you for raising this issue. Is this related to this?\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318", "2127811": "just pin this", "2123998": "Kudos for not giving up. Perseverance in the face of adversity and such ambiguity is a very admirable trait, especially adversity and ambiguity you brought upon yourselves by being transparent.\n\nI am really pulling for you folks that Kaggle does the right thing here. It would be a signal to folks who come later than being honest and open will be rewarded.", "2119743": "Believe that good will triumph over evil", "2127821": "Kaggle needs a credit system (Although I will be marked as a dishonest judgment debtors at that time🙈", "2119578": "It's all speculation without evidence. \nwhat evidence exists to prove that it's not participating on my own? ", "2138389": "Update. I've just found the answers here: https://www.kaggle.com/competitions/otto-recommender-system/discussion/381527 **\"How cheaters can steal tiers and reputations?\"**\n\n------\nI am new to competitions. Please explain to me why sharing code within a team makes a difference compared to sharing predictions. I see that the inability to share code was the cornerstone in this story.\nMy current understanding is as follows:\n- A winning team is required by the rules to submit the code that generated the winning submission, otherwise the team is disqualified. Sharing code within a team gives every team member confidence that their team can reproduce submissions. Thereby ensuring compliance with this particular part of the rules.\n- Sharing code within a team does _not_ guarantee at all that no cheating takes place. It is possible that an owner of multiple accounts takes code shared with them in one team and passes it on as their own to another team. Moreover, using many fake accounts the cheater can climb the ladder by joining successively stronger teams and end up in the top without even having compute power to execute the code they plagiarised. If the stolen code is not used by another winning team, they may even get away with their cheating.\n\nIs this correct? Am I missing anything? Thanks.", "2123339": "As a new-in-town in community, i was shocked that such unfair issues are raging and destroying this nice community. And you actually remind me of some very suspicious topics in \"XIAO HONG SHU\" saying that they guarantee sliver/bronze if you pay them....", "2123194": "It doesn't feel good to see kaggle's platform infiltrated by unfair mean", "2118744": "" } }