This paper is about this ongoing competition and uses the competition dataset.
\nIt seems it took large inspiration from @hengck23 Transformer architecture/code (including flattened images patching, ViT, the global Transfomer architecture, JiT acceleration etc…) without giving him any credit.
\nIt's a pity some researchers still take inspiration from kaggle competitions solutions, discussions and kernels without giving credit and some even downplay the importance of Kaggle for data science as emphasized by @cpmpml
\nEdit: The jit acceleration was introduced here by @affjljoo3581
", "messages": [ { "id": "1297274", "postDate": "05/07/2021 21:26:53", "content": "This paper is about this ongoing competition and uses the competition dataset.
\nIt seems it took large inspiration from @hengck23 Transformer architecture/code (including flattened images patching, ViT, the global Transfomer architecture, JiT acceleration etc…) without giving him any credit.
\nIt's a pity some researchers still take inspiration from kaggle competitions solutions, discussions and kernels without giving credit and some even downplay the importance of Kaggle for data science as emphasized by @cpmpml
\nEdit: The jit acceleration was introduced here by @affjljoo3581
", "rawMarkdown": "This [paper](https://arxiv.org/pdf/2104.14721.pdf) is about this ongoing competition and uses the competition dataset. \n \nIt seems it took large inspiration from @hengck23 Transformer architecture/code (including flattened images patching, ViT, the global Transfomer architecture, JiT acceleration etc...) without giving him any credit. \n\nIt's a pity some researchers still take inspiration from kaggle competitions solutions, discussions and kernels without giving credit and some even downplay the [importance of Kaggle](https://twitter.com/JFPuget/status/1320370798889230336) for data science as emphasized by @cpmpml \n\nEdit: The jit acceleration was introduced [here](https://www.kaggle.com/c/bms-molecular-translation/discussion/230498) by @affjljoo3581", "votes": null }, { "id": "1297286", "postDate": "05/07/2021 21:57:51", "content": "Wow, I didn't expect so direct theft. Did that really happen in the past with other competitions?
\nI guess @hengck23 or the Kaggle team could write to the University of Singapure and point that out to them.
Yes this already happened.
\nSee this tweet by CPMP for instance.
\nThis great solution was also \"stolen\" by some papers without credits.
\nCutmix was already used by @godaibo in his winning Kaggle solution way before the paper was published as emphasized by Jeremy Howard here . But in this case the authors didn't likely knew about his idea.
", "rawMarkdown": "Yes this already happened. \n\nSee this [tweet](https://twitter.com/JFPuget/status/1339993060130865155) by CPMP for instance. \n\n[This great solution ](https://www.kaggle.com/c/champs-scalar-coupling/discussion/106572) was also \"stolen\" by some papers without credits. \n\nCutmix was already used by @godaibo in his winning Kaggle solution way before the paper was published as emphasized by Jeremy Howard [here](https://twitter.com/jeremyphoward/status/1143364106893684737) . But in this case the authors didn't likely knew about his idea.", "votes": null }, { "id": "1297363", "postDate": "05/08/2021 00:37:04", "content": "This is terrible. Also the augmentation using RDKit sounds fake.
", "rawMarkdown": "This is terrible. Also the augmentation using RDKit sounds fake.", "votes": null }, { "id": "1297399", "postDate": "05/08/2021 01:35:30", "content": "Wow, I've never seen such a shameful paper. They don't even cite this competition dataset correctly.
\nAlso, even though they seems to evaluate their model's performance on the same train dataset, they model's performance is only 6.95 in Levenshtein Distance. That means they couldn't correctly reproduce discussion or notebook results😂
\nI can't call this as inspiration, purely plagiarism.
Jeremy Howard reads Kaggle… He was last seen 4 days ago here. I doubt he didn't see cutmix here first.
", "rawMarkdown": "Jeremy Howard reads Kaggle... He was last seen 4 days ago here. I doubt he didn't see cutmix here first.", "votes": null }, { "id": "1297621", "postDate": "05/08/2021 07:23:04", "content": "Seems like a desperate case of 'publish or perish' culture. They should have atleast given credits where credit is due.
", "rawMarkdown": "Seems like a desperate case of 'publish or perish' culture. They should have atleast given credits where credit is due.", "votes": null }, { "id": "1297746", "postDate": "05/08/2021 09:22:43", "content": "Interesting, that @hengck23 is also based in Singapore.
", "rawMarkdown": "Interesting, that @hengck23 is also based in Singapore.", "votes": null }, { "id": "1297756", "postDate": "05/08/2021 09:32:34", "content": "Yes, they only show an example of salt and pepper noise plus some erasing.
\nBut where are the things like: dropping an atom or converting double bonds to single bonds?
\nI doubt that they would not rather show those if they've really implemented them.
The paper is poorly written indeed. I'm not sure they've done all the experiments they claim
\nThere are a lot of sentences that are not well formulated. For instance here:
\n\n\nFollowing the success of Transformer in the field of natural language processing Vaswani et al. [2017]
\n
\n uses transformers in the decoder instead of RNN which results in an excellent performance
But Transformer was invented in this mentionned paper and for NLP field (Neural Machine Translation)
", "rawMarkdown": "The paper is poorly written indeed. I'm not sure they've done all the experiments they claim\n\nThere are a lot of sentences that are not well formulated. For instance here:\n> Following the success of Transformer in the field of natural language processing Vaswani et al. [2017]\nuses transformers in the decoder instead of RNN which results in an excellent performance\n\nBut Transformer was invented in this mentionned paper and for NLP field (Neural Machine Translation)", "votes": null }, { "id": "1298078", "postDate": "05/08/2021 14:37:46", "content": "by the way, jit acceleration was first introduced by me… 🙄🙄🙄
", "rawMarkdown": "by the way, jit acceleration was first introduced by me... 🙄🙄🙄", "votes": null }, { "id": "1299239", "postDate": "05/09/2021 15:03:18", "content": "I wonder how this is permitted given competition rules quoted below. Is this a matter for compliance?
\n''7. COMPETITION DATA.
\n\"Competition Data\" means the data or datasets available from the Competition Website for the purpose of use in the Competition, including any prototype or executable code provided on the Competition Website. The Competition Data will contain private and public test sets. Which data belongs to which set will not be made available to participants.
\nA. Data Access and Use. You may access and use the Competition Data only for participating in the Competition and on Kaggle.com forums. The Competition Sponsor reserves the right to disqualify any participant who uses the Competition Data other than as permitted by the Competition Website and these Rules.
\nB. Data Security. You agree to use reasonable and suitable measures to prevent persons who have not formally agreed to these Rules from gaining access to the Competition Data. You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition and you will be responsible for the same should they occur. You agree to notify Kaggle immediately upon learning of any possible unauthorized transmission of or unauthorized access to the Competition Data and agree to work with Kaggle to rectify any unauthorized transmission or access.\"
", "rawMarkdown": "I wonder how this is permitted given competition rules quoted below. Is this a matter for compliance? \n\n''7. COMPETITION DATA.\n\n\"Competition Data\" means the data or datasets available from the Competition Website for the purpose of use in the Competition, including any prototype or executable code provided on the Competition Website. The Competition Data will contain private and public test sets. Which data belongs to which set will not be made available to participants.\n\nA. Data Access and Use. You may access and use the Competition Data only for participating in the Competition and on Kaggle.com forums. The Competition Sponsor reserves the right to disqualify any participant who uses the Competition Data other than as permitted by the Competition Website and these Rules.\n\nB. Data Security. You agree to use reasonable and suitable measures to prevent persons who have not formally agreed to these Rules from gaining access to the Competition Data. You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition and you will be responsible for the same should they occur. You agree to notify Kaggle immediately upon learning of any possible unauthorized transmission of or unauthorized access to the Competition Data and agree to work with Kaggle to rectify any unauthorized transmission or access.\"", "votes": null }, { "id": "1301029", "postDate": "05/10/2021 22:54:10", "content": "In case people are wondering, here's the post by @hengck23. Also this post by @affjljoo3581 introduces the JIT torchscript strategy.
", "rawMarkdown": "In case people are wondering, here's [the post](https://www.kaggle.com/c/bms-molecular-translation/discussion/231190) by @hengck23. Also [this post](https://www.kaggle.com/c/bms-molecular-translation/discussion/230498) by @affjljoo3581 introduces the JIT torchscript strategy.", "votes": null }, { "id": "1305015", "postDate": "05/13/2021 04:16:59", "content": "omg… If this kind of thing keeps happening, I highly suspect if it will affect the motivation of sharing by kagglers.
", "rawMarkdown": "omg... If this kind of thing keeps happening, I highly suspect if it will affect the motivation of sharing by kagglers.", "votes": null }, { "id": "1348091", "postDate": "06/13/2021 18:10:42", "content": "\n\nJeremy Howard reads Kaggle… He was last seen 4 days ago here. I doubt he didn't see cutmix here first.
\n
Isn't that what he said in his tweet?
", "rawMarkdown": "> Jeremy Howard reads Kaggle… He was last seen 4 days ago here. I doubt he didn't see cutmix here first.\n\nIsn't that what he said in his tweet?", "votes": null } ], "comments": [ { "id": 1297286, "author_name": "cepheidq", "author_url": "", "post_date": "05/07/2021 21:57:51", "content": "Wow, I didn't expect so direct theft. Did that really happen in the past with other competitions?
\nI guess @hengck23 or the Kaggle team could write to the University of Singapure and point that out to them.
Yes this already happened.
\nSee this tweet by CPMP for instance.
\nThis great solution was also \"stolen\" by some papers without credits.
\nCutmix was already used by @godaibo in his winning Kaggle solution way before the paper was published as emphasized by Jeremy Howard here . But in this case the authors didn't likely knew about his idea.
", "votes": null, "replies": [] }, { "id": 1297591, "author_name": "cpmpml", "author_url": "", "post_date": "05/08/2021 06:41:27", "content": "Jeremy Howard reads Kaggle… He was last seen 4 days ago here. I doubt he didn't see cutmix here first.
", "votes": null, "replies": [] }, { "id": 1297746, "author_name": "talktocharles", "author_url": "", "post_date": "05/08/2021 09:22:43", "content": "Interesting, that @hengck23 is also based in Singapore.
", "votes": null, "replies": [] }, { "id": 1348091, "author_name": "agneev", "author_url": "", "post_date": "06/13/2021 18:10:42", "content": "\n\nJeremy Howard reads Kaggle… He was last seen 4 days ago here. I doubt he didn't see cutmix here first.
\n
Isn't that what he said in his tweet?
", "votes": null, "replies": [] } ] }, { "id": 1297363, "author_name": "houndcl", "author_url": "", "post_date": "05/08/2021 00:37:04", "content": "This is terrible. Also the augmentation using RDKit sounds fake.
", "votes": null, "replies": [ { "id": 1297756, "author_name": "nofreewill", "author_url": "", "post_date": "05/08/2021 09:32:34", "content": "Yes, they only show an example of salt and pepper noise plus some erasing.
\nBut where are the things like: dropping an atom or converting double bonds to single bonds?
\nI doubt that they would not rather show those if they've really implemented them.
Wow, I've never seen such a shameful paper. They don't even cite this competition dataset correctly.
\nAlso, even though they seems to evaluate their model's performance on the same train dataset, they model's performance is only 6.95 in Levenshtein Distance. That means they couldn't correctly reproduce discussion or notebook results😂
\nI can't call this as inspiration, purely plagiarism.
The paper is poorly written indeed. I'm not sure they've done all the experiments they claim
\nThere are a lot of sentences that are not well formulated. For instance here:
\n\n\nFollowing the success of Transformer in the field of natural language processing Vaswani et al. [2017]
\n
\n uses transformers in the decoder instead of RNN which results in an excellent performance
But Transformer was invented in this mentionned paper and for NLP field (Neural Machine Translation)
", "votes": null, "replies": [] } ] }, { "id": 1297621, "author_name": "pheadrus", "author_url": "", "post_date": "05/08/2021 07:23:04", "content": "Seems like a desperate case of 'publish or perish' culture. They should have atleast given credits where credit is due.
", "votes": null, "replies": [] }, { "id": 1298078, "author_name": "affjljoo3581", "author_url": "", "post_date": "05/08/2021 14:37:46", "content": "by the way, jit acceleration was first introduced by me… 🙄🙄🙄
", "votes": null, "replies": [] }, { "id": 1299239, "author_name": "something4kag", "author_url": "", "post_date": "05/09/2021 15:03:18", "content": "I wonder how this is permitted given competition rules quoted below. Is this a matter for compliance?
\n''7. COMPETITION DATA.
\n\"Competition Data\" means the data or datasets available from the Competition Website for the purpose of use in the Competition, including any prototype or executable code provided on the Competition Website. The Competition Data will contain private and public test sets. Which data belongs to which set will not be made available to participants.
\nA. Data Access and Use. You may access and use the Competition Data only for participating in the Competition and on Kaggle.com forums. The Competition Sponsor reserves the right to disqualify any participant who uses the Competition Data other than as permitted by the Competition Website and these Rules.
\nB. Data Security. You agree to use reasonable and suitable measures to prevent persons who have not formally agreed to these Rules from gaining access to the Competition Data. You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition and you will be responsible for the same should they occur. You agree to notify Kaggle immediately upon learning of any possible unauthorized transmission of or unauthorized access to the Competition Data and agree to work with Kaggle to rectify any unauthorized transmission or access.\"
", "votes": null, "replies": [] }, { "id": 1301029, "author_name": "xhlulu", "author_url": "", "post_date": "05/10/2021 22:54:10", "content": "In case people are wondering, here's the post by @hengck23. Also this post by @affjljoo3581 introduces the JIT torchscript strategy.
", "votes": null, "replies": [] }, { "id": 1305015, "author_name": "haqishen", "author_url": "", "post_date": "05/13/2021 04:16:59", "content": "omg… If this kind of thing keeps happening, I highly suspect if it will affect the motivation of sharing by kagglers.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1297274": "This [paper](https://arxiv.org/pdf/2104.14721.pdf) is about this ongoing competition and uses the competition dataset. \n \nIt seems it took large inspiration from @hengck23 Transformer architecture/code (including flattened images patching, ViT, the global Transfomer architecture, JiT acceleration etc...) without giving him any credit. \n\nIt's a pity some researchers still take inspiration from kaggle competitions solutions, discussions and kernels without giving credit and some even downplay the [importance of Kaggle](https://twitter.com/JFPuget/status/1320370798889230336) for data science as emphasized by @cpmpml \n\nEdit: The jit acceleration was introduced [here](https://www.kaggle.com/c/bms-molecular-translation/discussion/230498) by @affjljoo3581", "1297286": "Wow, I didn't expect so direct theft. Did that really happen in the past with other competitions?\nI guess @hengck23 or the Kaggle team could write to the [University of Singapure](https://www.nus.edu.sg/) and point that out to them.", "1297303": "Yes this already happened. \n\nSee this [tweet](https://twitter.com/JFPuget/status/1339993060130865155) by CPMP for instance. \n\n[This great solution ](https://www.kaggle.com/c/champs-scalar-coupling/discussion/106572) was also \"stolen\" by some papers without credits. \n\nCutmix was already used by @godaibo in his winning Kaggle solution way before the paper was published as emphasized by Jeremy Howard [here](https://twitter.com/jeremyphoward/status/1143364106893684737) . But in this case the authors didn't likely knew about his idea.", "1297363": "This is terrible. Also the augmentation using RDKit sounds fake.", "1297399": "Wow, I've never seen such a shameful paper. They don't even cite this competition dataset correctly.\nAlso, even though they seems to evaluate their model's performance on **the same train dataset**, they model's performance is only 6.95 in Levenshtein Distance. That means they couldn't correctly reproduce discussion or notebook results😂\nI can't call this as inspiration, purely **plagiarism**.", "1297591": "Jeremy Howard reads Kaggle... He was last seen 4 days ago here. I doubt he didn't see cutmix here first.", "1297621": "Seems like a desperate case of 'publish or perish' culture. They should have atleast given credits where credit is due.", "1297746": "Interesting, that @hengck23 is also based in Singapore.", "1297756": "Yes, they only show an example of salt and pepper noise plus some erasing.\nBut where are the things like: `dropping an atom` or `converting double bonds to single bonds`?\nI doubt that they would not rather show those if they've really implemented them.", "1298063": "The paper is poorly written indeed. I'm not sure they've done all the experiments they claim\n\nThere are a lot of sentences that are not well formulated. For instance here:\n> Following the success of Transformer in the field of natural language processing Vaswani et al. [2017]\nuses transformers in the decoder instead of RNN which results in an excellent performance\n\nBut Transformer was invented in this mentionned paper and for NLP field (Neural Machine Translation)", "1298078": "by the way, jit acceleration was first introduced by me... 🙄🙄🙄", "1299239": "I wonder how this is permitted given competition rules quoted below. Is this a matter for compliance? \n\n''7. COMPETITION DATA.\n\n\"Competition Data\" means the data or datasets available from the Competition Website for the purpose of use in the Competition, including any prototype or executable code provided on the Competition Website. The Competition Data will contain private and public test sets. Which data belongs to which set will not be made available to participants.\n\nA. Data Access and Use. You may access and use the Competition Data only for participating in the Competition and on Kaggle.com forums. The Competition Sponsor reserves the right to disqualify any participant who uses the Competition Data other than as permitted by the Competition Website and these Rules.\n\nB. Data Security. You agree to use reasonable and suitable measures to prevent persons who have not formally agreed to these Rules from gaining access to the Competition Data. You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition and you will be responsible for the same should they occur. You agree to notify Kaggle immediately upon learning of any possible unauthorized transmission of or unauthorized access to the Competition Data and agree to work with Kaggle to rectify any unauthorized transmission or access.\"", "1301029": "In case people are wondering, here's [the post](https://www.kaggle.com/c/bms-molecular-translation/discussion/231190) by @hengck23. Also [this post](https://www.kaggle.com/c/bms-molecular-translation/discussion/230498) by @affjljoo3581 introduces the JIT torchscript strategy.", "1305015": "omg... If this kind of thing keeps happening, I highly suspect if it will affect the motivation of sharing by kagglers.", "1348091": "> Jeremy Howard reads Kaggle… He was last seen 4 days ago here. I doubt he didn't see cutmix here first.\n\nIsn't that what he said in his tweet?" }, "source": "meta" }