On the competition rules page, under Data Security, it states \"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. 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.\"
\n@deeppast I am not sure how to interpret this in the context of using LLMs to interact or understand the competition datasets. I know the rules does not forbid the use of LLMs. However, it appears to me that using models like Gemini with competition data, would constitute permitting persons who have not agreed to the rules, access to competition data.
\nAn example would be the use of online services for translating some of the non-English translations in the publications.csv. Ryan's response in this thread only seem to relate to the generation of synthetic data, whereas my question regards allowing access to the competition data. What is the recommended way to translate some of these translations that are in German, French, etc without allowing access to third party applications?
On the competition rules page, under Data Security, it states \"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. 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.\"
\n@deeppast I am not sure how to interpret this in the context of using LLMs to interact or understand the competition datasets. I know the rules does not forbid the use of LLMs. However, it appears to me that using models like Gemini with competition data, would constitute permitting persons who have not agreed to the rules, access to competition data.
\nAn example would be the use of online services for translating some of the non-English translations in the publications.csv. Ryan's response in this thread only seem to relate to the generation of synthetic data, whereas my question regards allowing access to the competition data. What is the recommended way to translate some of these translations that are in German, French, etc without allowing access to third party applications?
You may find this comment by DeepPast of some use.
\nI asked a similar question on text processing using externally sourced LLM models.
", "rawMarkdown": "You may find [this comment by DeepPast](https://www.kaggle.com/competitions/deep-past-initiative-machine-translation/discussion/664775#3382894) of some use.\n\nI asked a similar question on text processing using externally sourced LLM models.", "votes": null }, { "id": "3387598", "postDate": "01/07/2026 11:06:50", "content": "Thank you for the link. I have seen their response, but it seems to me not clear enough. When they say \"… advisable to keep the use of this data for ML and not public display, and for any models as outcomes to cite the bibliography we provided for continuity.\", they don't seem to explicitly speak on the use of external LLMs for processing.
\nYour response to their comment seem to indicate that you're unsure of what to make of their comment. Let me know if you have anymore insights on this.
", "rawMarkdown": "Thank you for the link. I have seen their response, but it seems to me not clear enough. When they say \"... advisable to keep the use of this data for ML and not public display, and for any models as outcomes to cite the bibliography we provided for continuity.\", they don't seem to explicitly speak on the use of external LLMs for processing.\n\nYour response to their comment seem to indicate that you're unsure of what to make of their comment. Let me know if you have anymore insights on this.", "votes": null }, { "id": "3387812", "postDate": "01/07/2026 16:56:35", "content": "I don't. I think that generally speaking, the data processed in this competition should only be processed for the purpose of the Competition. If so, any processing by a third party should be excluded as we cannot guarantee that it is not copied while being processed on the external servers. Enterprise Tiers sometimes give that kind of guarantee but that's a pinky promise and cannot be verified on the user's end.
\nSo, my take on it is that we can use LLM as much as we like, so long as we:
\nBut for official statement we should wait for the Hosts' response.
", "rawMarkdown": "I don't. I think that generally speaking, the data processed in this competition should only be processed for the purpose of the Competition. If so, any processing by a third party should be excluded as we cannot guarantee that it is not copied while being processed on the external servers. Enterprise Tiers sometimes give that kind of guarantee but that's a pinky promise and cannot be verified on the user's end.\n\nSo, my take on it is that we can use LLM as much as we like, so long as we:\n1. process the data while making sure it does not leak beyond our control (local LLM training, fine-tuning, etc.)\n2. make our end results public and cite the source material.\n\nBut for official statement we should wait for the Hosts' response.", "votes": null }, { "id": "3387907", "postDate": "01/07/2026 19:18:59", "content": "Thank you. I agree with your observation that it would generally be a violation of the rules to involve an LLM owned by a third party for any form of processing.
\nIndeed! I'm curious to hear the host's opinion on this.
", "rawMarkdown": "Thank you. I agree with your observation that it would generally be a violation of the rules to involve an LLM owned by a third party for any form of processing.\n\nIndeed! I'm curious to hear the host's opinion on this.", "votes": null }, { "id": "3388804", "postDate": "01/09/2026 16:07:28", "content": "I found this response from the host: https://www.kaggle.com/competitions/deep-past-initiative-machine-translation/discussion/663656#3380517
", "rawMarkdown": "I found this response from the host: https://www.kaggle.com/competitions/deep-past-initiative-machine-translation/discussion/663656#3380517", "votes": null } ], "comments": [ { "id": 3387588, "author_name": "angantyr", "author_url": "", "post_date": "01/07/2026 10:49:35", "content": "You may find this comment by DeepPast of some use.
\nI asked a similar question on text processing using externally sourced LLM models.
", "votes": null, "replies": [ { "id": 3387598, "author_name": "anaphase21", "author_url": "", "post_date": "01/07/2026 11:06:50", "content": "Thank you for the link. I have seen their response, but it seems to me not clear enough. When they say \"… advisable to keep the use of this data for ML and not public display, and for any models as outcomes to cite the bibliography we provided for continuity.\", they don't seem to explicitly speak on the use of external LLMs for processing.
\nYour response to their comment seem to indicate that you're unsure of what to make of their comment. Let me know if you have anymore insights on this.
", "votes": null, "replies": [ { "id": 3387812, "author_name": "angantyr", "author_url": "", "post_date": "01/07/2026 16:56:35", "content": "I don't. I think that generally speaking, the data processed in this competition should only be processed for the purpose of the Competition. If so, any processing by a third party should be excluded as we cannot guarantee that it is not copied while being processed on the external servers. Enterprise Tiers sometimes give that kind of guarantee but that's a pinky promise and cannot be verified on the user's end.
\nSo, my take on it is that we can use LLM as much as we like, so long as we:
\nBut for official statement we should wait for the Hosts' response.
", "votes": null, "replies": [ { "id": 3387907, "author_name": "anaphase21", "author_url": "", "post_date": "01/07/2026 19:18:59", "content": "Thank you. I agree with your observation that it would generally be a violation of the rules to involve an LLM owned by a third party for any form of processing.
\nIndeed! I'm curious to hear the host's opinion on this.
", "votes": null, "replies": [] }, { "id": 3388804, "author_name": "anaphase21", "author_url": "", "post_date": "01/09/2026 16:07:28", "content": "I found this response from the host: https://www.kaggle.com/competitions/deep-past-initiative-machine-translation/discussion/663656#3380517
", "votes": null, "replies": [] } ] } ] } ] } ], "raw_markdown_by_id": { "3387512": "On the competition rules page, under Data Security, it states *\"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. 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.\"*\n\n@deeppast I am not sure how to interpret this in the context of using LLMs to interact or understand the competition datasets. I know the rules does not forbid the use of LLMs. However, it appears to me that using models like Gemini with competition data, would constitute permitting persons who have not agreed to the rules, access to competition data.\n\nAn example would be the use of online services for translating some of the non-English translations in the `publications.csv`. Ryan's response in this [thread](https://www.kaggle.com/competitions/deep-past-initiative-machine-translation/discussion/663656#3380517) only seem to relate to the generation of synthetic data, whereas my question regards allowing access to the competition data. What is the recommended way to translate some of these translations that are in German, French, etc without allowing access to third party applications?", "3387588": "You may find [this comment by DeepPast](https://www.kaggle.com/competitions/deep-past-initiative-machine-translation/discussion/664775#3382894) of some use.\n\nI asked a similar question on text processing using externally sourced LLM models.", "3387598": "Thank you for the link. I have seen their response, but it seems to me not clear enough. When they say \"... advisable to keep the use of this data for ML and not public display, and for any models as outcomes to cite the bibliography we provided for continuity.\", they don't seem to explicitly speak on the use of external LLMs for processing.\n\nYour response to their comment seem to indicate that you're unsure of what to make of their comment. Let me know if you have anymore insights on this.", "3387812": "I don't. I think that generally speaking, the data processed in this competition should only be processed for the purpose of the Competition. If so, any processing by a third party should be excluded as we cannot guarantee that it is not copied while being processed on the external servers. Enterprise Tiers sometimes give that kind of guarantee but that's a pinky promise and cannot be verified on the user's end.\n\nSo, my take on it is that we can use LLM as much as we like, so long as we:\n1. process the data while making sure it does not leak beyond our control (local LLM training, fine-tuning, etc.)\n2. make our end results public and cite the source material.\n\nBut for official statement we should wait for the Hosts' response.", "3387907": "Thank you. I agree with your observation that it would generally be a violation of the rules to involve an LLM owned by a third party for any form of processing.\n\nIndeed! I'm curious to hear the host's opinion on this.", "3388804": "I found this response from the host: https://www.kaggle.com/competitions/deep-past-initiative-machine-translation/discussion/663656#3380517" }, "source": "meta" }