{
  "id": 678974,
  "title": "Mid-competition updates have now become a trend in Kaggle competitions.",
  "url": "/competitions/deep-past-initiative-machine-translation/discussion/678974",
  "author_name": "",
  "post_date": "2026-02-26T07:20:11.986247100Z",
  "votes": 25,
  "comment_count": 6,
  "views": 0,
  "content": "<p>The NFL2026 competition I participated in recently experienced a mid-process update, and now this DPC competition is doing the same. </p>\n<p>I've also noticed that the Vesuvius Challenge has experienced multiple mid-process updates. </p>\n<p>What's going on with Kaggle? 😅</p>",
  "messages": [
    {
      "id": "3414185",
      "postDate": "02/26/2026 07:20:11",
      "content": "<p>The NFL2026 competition I participated in recently experienced a mid-process update, and now this DPC competition is doing the same. </p>\n<p>I've also noticed that the Vesuvius Challenge has experienced multiple mid-process updates. </p>\n<p>What's going on with Kaggle? 😅</p>",
      "rawMarkdown": "The NFL2026 competition I participated in recently experienced a mid-process update, and now this DPC competition is doing the same. \n\nI've also noticed that the Vesuvius Challenge has experienced multiple mid-process updates. \n\nWhat's going on with Kaggle? 😅",
      "votes": null
    },
    {
      "id": "3414188",
      "postDate": "02/26/2026 07:30:41",
      "content": "<p>It’s the new tactic to prevent overfitting </p>",
      "rawMarkdown": "It’s the new tactic to prevent overfitting",
      "votes": null
    },
    {
      "id": "3414189",
      "postDate": "02/26/2026 07:57:17",
      "content": "<p>My score got decreased from 34.9 to 34.2 but rank increased from 552 to 124 what is happening?</p>",
      "rawMarkdown": "My score got decreased from 34.9 to 34.2 but rank increased from 552 to 124 what is happening?",
      "votes": null
    },
    {
      "id": "3414265",
      "postDate": "02/26/2026 11:05:58",
      "content": "<p>I think it's because of LLMs. At the age before ChatGPT, the hosts have to decide everything themselves. So they had to make every decision carefully. </p>\n<p>Now many decisions may actually be made by AI, instead of the hosts themselves. it may be much easier to host, but lead to carelessness.</p>",
      "rawMarkdown": "I think it's because of LLMs. At the age before ChatGPT, the hosts have to decide everything themselves. So they had to make every decision carefully. \n\nNow many decisions may actually be made by AI, instead of the hosts themselves. it may be much easier to host, but lead to carelessness.",
      "votes": null
    },
    {
      "id": "3414555",
      "postDate": "02/27/2026 05:37:05",
      "content": "<p>Kaggle being Kaggle, I think <a href=\"https://www.kaggle.com/hengzheng\" target=\"_blank\">@hengzheng</a> </p>",
      "rawMarkdown": "Kaggle being Kaggle, I think @hengzheng",
      "votes": null
    },
    {
      "id": "3416794",
      "postDate": "03/03/2026 18:56:24",
      "content": "<p>complexity?</p>",
      "rawMarkdown": "complexity?",
      "votes": null
    },
    {
      "id": "3416833",
      "postDate": "03/03/2026 20:45:50",
      "content": "<p>Yes, and it has really helped to burn me out faster. We were in the lead for the Vesuvius comp. Tested hundreds of experiments locally and then with 4 days remaining they change the data and add two weeks. The change gave us not enough time to retest or understand the data but plenty of time for us to lose our place. Still happy with the result and more importantly I learned a lot but that would have been really cool for us. On top of that I know the prize would be equivalent to many months of salary for some of my teammates. </p>\n<p>I understand wishing you did things differently and even adjusting things in early comp to correct for mistakes. But for the sake of competition, it just seems like bad practice that reduces how great the platform is. If it's possible you would think to take the extra few weeks to get the data and the metric right. </p>",
      "rawMarkdown": "Yes, and it has really helped to burn me out faster. We were in the lead for the Vesuvius comp. Tested hundreds of experiments locally and then with 4 days remaining they change the data and add two weeks. The change gave us not enough time to retest or understand the data but plenty of time for us to lose our place. Still happy with the result and more importantly I learned a lot but that would have been really cool for us. On top of that I know the prize would be equivalent to many months of salary for some of my teammates. \n\nI understand wishing you did things differently and even adjusting things in early comp to correct for mistakes. But for the sake of competition, it just seems like bad practice that reduces how great the platform is. If it's possible you would think to take the extra few weeks to get the data and the metric right.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3414188,
      "author_name": "einherjer",
      "author_url": "",
      "post_date": "02/26/2026 07:30:41",
      "content": "<p>It’s the new tactic to prevent overfitting </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3414189,
      "author_name": "adarsh2626",
      "author_url": "",
      "post_date": "02/26/2026 07:57:17",
      "content": "<p>My score got decreased from 34.9 to 34.2 but rank increased from 552 to 124 what is happening?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3414265,
      "author_name": "forcewithme",
      "author_url": "",
      "post_date": "02/26/2026 11:05:58",
      "content": "<p>I think it's because of LLMs. At the age before ChatGPT, the hosts have to decide everything themselves. So they had to make every decision carefully. </p>\n<p>Now many decisions may actually be made by AI, instead of the hosts themselves. it may be much easier to host, but lead to carelessness.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3414555,
      "author_name": "navneetbende",
      "author_url": "",
      "post_date": "02/27/2026 05:37:05",
      "content": "<p>Kaggle being Kaggle, I think <a href=\"https://www.kaggle.com/hengzheng\" target=\"_blank\">@hengzheng</a> </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3416794,
      "author_name": "davidochoacorrales",
      "author_url": "",
      "post_date": "03/03/2026 18:56:24",
      "content": "<p>complexity?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3416833,
      "author_name": "cody11null",
      "author_url": "",
      "post_date": "03/03/2026 20:45:50",
      "content": "<p>Yes, and it has really helped to burn me out faster. We were in the lead for the Vesuvius comp. Tested hundreds of experiments locally and then with 4 days remaining they change the data and add two weeks. The change gave us not enough time to retest or understand the data but plenty of time for us to lose our place. Still happy with the result and more importantly I learned a lot but that would have been really cool for us. On top of that I know the prize would be equivalent to many months of salary for some of my teammates. </p>\n<p>I understand wishing you did things differently and even adjusting things in early comp to correct for mistakes. But for the sake of competition, it just seems like bad practice that reduces how great the platform is. If it's possible you would think to take the extra few weeks to get the data and the metric right. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3414185": "The NFL2026 competition I participated in recently experienced a mid-process update, and now this DPC competition is doing the same. \n\nI've also noticed that the Vesuvius Challenge has experienced multiple mid-process updates. \n\nWhat's going on with Kaggle? 😅",
    "3414188": "It’s the new tactic to prevent overfitting",
    "3414189": "My score got decreased from 34.9 to 34.2 but rank increased from 552 to 124 what is happening?",
    "3414265": "I think it's because of LLMs. At the age before ChatGPT, the hosts have to decide everything themselves. So they had to make every decision carefully. \n\nNow many decisions may actually be made by AI, instead of the hosts themselves. it may be much easier to host, but lead to carelessness.",
    "3414555": "Kaggle being Kaggle, I think @hengzheng",
    "3416794": "complexity?",
    "3416833": "Yes, and it has really helped to burn me out faster. We were in the lead for the Vesuvius comp. Tested hundreds of experiments locally and then with 4 days remaining they change the data and add two weeks. The change gave us not enough time to retest or understand the data but plenty of time for us to lose our place. Still happy with the result and more importantly I learned a lot but that would have been really cool for us. On top of that I know the prize would be equivalent to many months of salary for some of my teammates. \n\nI understand wishing you did things differently and even adjusting things in early comp to correct for mistakes. But for the sake of competition, it just seems like bad practice that reduces how great the platform is. If it's possible you would think to take the extra few weeks to get the data and the metric right."
  },
  "source": "meta"
}