![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F11991501%2F19e34aff4dcde4da5c9ed81991a685fb%2F2023-06-23%2019.32.26.png?generation=1687516407815692&alt=media\")
https://www.kaggle.com/competitions/open-problems-multimodal/discussion/366313
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381495
I hope our efforts will be properly rewarded.
", "messages": [ { "id": 2313328, "postDate": "2023-06-22T15:31:23.230Z", "content": "https://www.kaggle.com/competitions/open-problems-multimodal/discussion/366313
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381495
I hope our efforts will be properly rewarded.
", "rawMarkdown": "https://www.kaggle.com/competitions/open-problems-multimodal/discussion/366313\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381495\n\nI hope our efforts will be properly rewarded.", "votes": 17 }, { "id": 2313469, "postDate": "2023-06-22T17:44:31.123Z", "content": "I guess also for this competition someone kept posting high score notebook. Some trivial change in hyperparameters/blending based on that would boost the public score. However we could expect some shake in private LB if many teams are using extremely similar setup & if that setup is not robust.
", "rawMarkdown": "I guess also for this competition someone kept posting high score notebook. Some trivial change in hyperparameters/blending based on that would boost the public score. However we could expect some shake in private LB if many teams are using extremely similar setup & if that setup is not robust. ", "votes": 3 }, { "id": 2313358, "postDate": "2023-06-22T15:57:23.020Z", "content": "Yes, you are right. I moved from about 80th to 120th in a few days, it's too strange. I think Kaggle should take action to stop these behaviors. Perhaps Kaggle can host a competition called Kaggle User Fraud Detection.
", "rawMarkdown": "Yes, you are right. I moved from about 80th to 120th in a few days, it's too strange. I think Kaggle should take action to stop these behaviors. Perhaps Kaggle can host a competition called Kaggle User Fraud Detection.", "votes": 3, "replies": [ { "id": 2313885, "postDate": "2023-06-23T02:45:19.947Z", "content": "The same, I am from ~8x to 115 in just a few days. Every time I check the leaderboard, I drop 1-2 rank, and there are some new teams join.
", "rawMarkdown": "The same, I am from ~8x to 115 in just a few days. Every time I check the leaderboard, I drop 1-2 rank, and there are some new teams join.", "votes": 2 }, { "id": 2314920, "postDate": "2023-06-23T17:12:32.017Z", "content": "That's quite common, especially for large competitions. Many vicious combatants enjoy holding their submissions till the last second to sap the unsuspected prey out of the silver zone.
", "rawMarkdown": "That's quite common, especially for large competitions. Many vicious combatants enjoy holding their submissions till the last second to sap the unsuspected prey out of the silver zone. " } ] }, { "id": 2314876, "postDate": "2023-06-23T16:47:24.103Z", "content": "Indeed, I also noticed some people selling gold/silver solutions on XiaoHongShu (Chinese Twitter).
\nBut the sad reality is, I'm not sure Kaggle can help with this. It's really easy to make the model produce slightly different scores by e.g. training it one more epoch or having one more booster. The inference code can simply be written by each individual buyer with minimal effort, to create variation in the submitted code. Reporting the post wouldn't really help either, even if the account gets banned, the seller can just create another account.
", "rawMarkdown": "Indeed, I also noticed some people selling gold/silver solutions on XiaoHongShu (Chinese Twitter).\n\nBut the sad reality is, I'm not sure Kaggle can help with this. It's really easy to make the model produce slightly different scores by e.g. training it one more epoch or having one more booster. The inference code can simply be written by each individual buyer with minimal effort, to create variation in the submitted code. Reporting the post wouldn't really help either, even if the account gets banned, the seller can just create another account.", "votes": 4, "replies": [ { "id": 2314890, "postDate": "2023-06-23T16:55:38.307Z", "content": "Yes, code competitions don't really prevent much actually, as only the inference code, which is the easier part, is visible.
", "rawMarkdown": "Yes, code competitions don't really prevent much actually, as only the inference code, which is the easier part, is visible.", "votes": 1, "replies": [ { "id": 2314906, "postDate": "2023-06-23T17:05:00.880Z", "content": "Indeed. But I think the ones who cheat are in the end not the majority, what options do we have other than believe in the decency within each other 🙃🙃
", "rawMarkdown": "Indeed. But I think the ones who cheat are in the end not the majority, what options do we have other than believe in the decency within each other 🙃🙃" }, { "id": 2314911, "postDate": "2023-06-23T17:10:15.877Z", "content": "according to this comment it seems they don't even need to change the models
", "rawMarkdown": "according to this [comment](https://www.kaggle.com/competitions/predict-student-performance-from-game-play/discussion/418802#2314319) it seems they don't even need to change the models ", "votes": 2 }, { "id": 2314925, "postDate": "2023-06-23T17:18:12.587Z", "content": "Well… At least you guys at the top are safe from this bs, only us silver plebs need to worry about these issues 😒
", "rawMarkdown": "Well... At least you guys at the top are safe from this bs, only us silver plebs need to worry about these issues 😒", "votes": 2 }, { "id": 2314928, "postDate": "2023-06-23T17:21:05.820Z", "content": "haha everything can turn upside down in the end 😄
", "rawMarkdown": "haha everything can turn upside down in the end 😄", "votes": 3 }, { "id": 2315175, "postDate": "2023-06-23T21:53:14.017Z", "content": "Yes, but for the medal range Kaggle could also require submission of the fitting code.
", "rawMarkdown": "Yes, but for the medal range Kaggle could also require submission of the fitting code.", "votes": 2 }, { "id": 2315291, "postDate": "2023-06-24T03:11:16.697Z", "content": "@kononenko, that would be a good idea
", "rawMarkdown": "@kononenko, that would be a good idea", "votes": 1 } ] } ] }, { "id": 2314273, "postDate": "2023-06-23T10:10:52.963Z", "content": "Since this is a code competition, it shouldn’t be difficult to detect private notebook sharing. Hopefully Kaggle team will take care of.
", "rawMarkdown": "Since this is a code competition, it shouldn’t be difficult to detect private notebook sharing. Hopefully Kaggle team will take care of.", "votes": 1, "replies": [ { "id": 2314319, "postDate": "2023-06-23T10:41:31.797Z", "content": "In fact, I don't think there are any expectations.
\nAMP-Parkinson competition was a code competition, but even these teams were not excluded.
\nOf course, I understand that they could happen to be teams of 4 or 5, happen to have the same score, and happen to be in the same rise range.
\nCan't kaggle team just ask some suspicious teams for the code? Since there are not that many of them, it is not a big hassle.Most users will gladly send you the codes.
https://www.kaggle.com/competitions/amp-parkinsons-disease-progression-prediction/leaderboard
\n
I guess also for this competition someone kept posting high score notebook. Some trivial change in hyperparameters/blending based on that would boost the public score. However we could expect some shake in private LB if many teams are using extremely similar setup & if that setup is not robust.
", "votes": 3, "replies": [] }, { "id": 2313358, "author_name": "Jiexi Zhang", "author_url": "", "post_date": "2023-06-22T15:57:23.020000", "content": "Yes, you are right. I moved from about 80th to 120th in a few days, it's too strange. I think Kaggle should take action to stop these behaviors. Perhaps Kaggle can host a competition called Kaggle User Fraud Detection.
", "votes": 3, "replies": [ { "id": 2313885, "author_name": "Lam Pham Ba Tung", "author_url": "", "post_date": "2023-06-23T02:45:19.947000", "content": "The same, I am from ~8x to 115 in just a few days. Every time I check the leaderboard, I drop 1-2 rank, and there are some new teams join.
", "votes": 2, "replies": [] }, { "id": 2314920, "author_name": "Woprime", "author_url": "", "post_date": "2023-06-23T17:12:32.017000", "content": "That's quite common, especially for large competitions. Many vicious combatants enjoy holding their submissions till the last second to sap the unsuspected prey out of the silver zone.
", "votes": 0, "replies": [] } ] }, { "id": 2314876, "author_name": "Woprime", "author_url": "", "post_date": "2023-06-23T16:47:24.103000", "content": "Indeed, I also noticed some people selling gold/silver solutions on XiaoHongShu (Chinese Twitter).
\nBut the sad reality is, I'm not sure Kaggle can help with this. It's really easy to make the model produce slightly different scores by e.g. training it one more epoch or having one more booster. The inference code can simply be written by each individual buyer with minimal effort, to create variation in the submitted code. Reporting the post wouldn't really help either, even if the account gets banned, the seller can just create another account.
", "votes": 4, "replies": [ { "id": 2314890, "author_name": "Elias", "author_url": "", "post_date": "2023-06-23T16:55:38.307000", "content": "Yes, code competitions don't really prevent much actually, as only the inference code, which is the easier part, is visible.
", "votes": 1, "replies": [ { "id": 2314906, "author_name": "Woprime", "author_url": "", "post_date": "2023-06-23T17:05:00.880000", "content": "Indeed. But I think the ones who cheat are in the end not the majority, what options do we have other than believe in the decency within each other 🙃🙃
", "votes": 0, "replies": [] }, { "id": 2314911, "author_name": "empty", "author_url": "", "post_date": "2023-06-23T17:10:15.877000", "content": "according to this comment it seems they don't even need to change the models
", "votes": 2, "replies": [] }, { "id": 2314925, "author_name": "Woprime", "author_url": "", "post_date": "2023-06-23T17:18:12.587000", "content": "Well… At least you guys at the top are safe from this bs, only us silver plebs need to worry about these issues 😒
", "votes": 2, "replies": [] }, { "id": 2314928, "author_name": "empty", "author_url": "", "post_date": "2023-06-23T17:21:05.820000", "content": "haha everything can turn upside down in the end 😄
", "votes": 3, "replies": [] }, { "id": 2315175, "author_name": "Oleksiy Kononenko", "author_url": "", "post_date": "2023-06-23T21:53:14.017000", "content": "Yes, but for the medal range Kaggle could also require submission of the fitting code.
", "votes": 2, "replies": [] }, { "id": 2315291, "author_name": "Elias", "author_url": "", "post_date": "2023-06-24T03:11:16.697000", "content": "@kononenko, that would be a good idea
", "votes": 1, "replies": [] } ] } ] }, { "id": 2314273, "author_name": "Oleksiy Kononenko", "author_url": "", "post_date": "2023-06-23T10:10:52.963000", "content": "Since this is a code competition, it shouldn’t be difficult to detect private notebook sharing. Hopefully Kaggle team will take care of.
", "votes": 1, "replies": [ { "id": 2314319, "author_name": "wakaka", "author_url": "", "post_date": "2023-06-23T10:41:31.797000", "content": "In fact, I don't think there are any expectations.
\nAMP-Parkinson competition was a code competition, but even these teams were not excluded.
\nOf course, I understand that they could happen to be teams of 4 or 5, happen to have the same score, and happen to be in the same rise range.
\nCan't kaggle team just ask some suspicious teams for the code? Since there are not that many of them, it is not a big hassle.Most users will gladly send you the codes.
https://www.kaggle.com/competitions/amp-parkinsons-disease-progression-prediction/leaderboard
\n