{
  "topic": {
    "id": 736268,
    "title": "Too many noise public notebookS.",
    "authorName": "XuKong Ji",
    "commentCount": 13,
    "votes": 13,
    "postDate": "2026-08-20T04:31:34.063000"
  },
  "comments": [
    {
      "id": 3518150,
      "authorName": "Tucker Arrants",
      "votes": 3,
      "postDate": "2026-08-29T16:57:18.110000",
      "content": "<p>You can also just check for notebooks with comments. Notebooks with lots of likes and no comments are just high scoring, overfit to LB, AI generated noise. This alone can find the few useful publicly shared notebooks. </p>"
    },
    {
      "id": 3514799,
      "authorName": "PC Jimmmy",
      "votes": 1,
      "postDate": "2026-08-20T06:13:14.623000",
      "content": "<p>Sadly the public notebooks has become the place were you fork and hope to get some up votes.</p>\n<p>I do sort by recently created and NEVER look at folks without a decent kaggle ranking as I plan my what the heck am I going to try today list.</p>"
    },
    {
      "id": 3515244,
      "authorName": "XuKong Ji",
      "votes": 1,
      "postDate": "2026-08-21T07:36:14.320000",
      "content": "<p>Totally agree. Too many notebooks are just ensembling spam for GM points. </p>"
    },
    {
      "id": 3514969,
      "authorName": "Tucker Arrants",
      "votes": 2,
      "postDate": "2026-08-20T13:55:36.063000",
      "content": "<p><a href=\"https://www.kaggle.com/competitions/rogii-wellbore-geology-prediction/discussion/718670\" target=\"_blank\">This discussion from ROGII-Wellbore</a> might be useful to sift through the noise of public notebooks. </p>"
    },
    {
      "id": 3515246,
      "authorName": "XuKong Ji",
      "votes": 1,
      "postDate": "2026-08-21T07:41:22.523000",
      "content": "<p>Thanks for share this. Will check it out</p>"
    },
    {
      "id": 3518140,
      "authorName": "Homurraa",
      "votes": 0,
      "postDate": "2026-08-29T16:26:25.403000",
      "content": "<p>This is what happens at the coding agent era. Generating lots of contents (maybe low quality ones) is cheap</p>"
    },
    {
      "id": 3514970,
      "authorName": "Ahmet Kiraz",
      "votes": 0,
      "postDate": "2026-08-20T13:57:10.140000",
      "content": "<p>A project that’s a bit of a challenge for me, but one that benefits humanity…</p>"
    },
    {
      "id": 3514915,
      "authorName": "Mattia Angeli",
      "votes": 0,
      "postDate": "2026-08-20T11:57:46.930000",
      "content": "<p>It’s currently a race to overfit the public leaderboard, 0.001 at a time 😅</p>"
    },
    {
      "id": 3514926,
      "authorName": "Mattia Angeli",
      "votes": 4,
      "postDate": "2026-08-20T12:25:37.590000",
      "content": "<p>Here's the only cell you need from several dozen top-scoring public notebooks:</p>\n<p><strong><a href=\"https://www.kaggle.com/code/mattiaangeli/bend-the-knee-to-dinov3-ensembled?scriptVersionId=343218600&amp;cellId=24\" target=\"_blank\">DINOv3 architecture</a></strong></p>\n<p>It was only a modest addition to <a href=\"https://www.kaggle.com/datasets/pilkwang/rsna-knee-weights\" target=\"_blank\">Pilkwang's 20-model DINOv2 ensemble</a>, since their errors remain fairly correlated.</p>"
    },
    {
      "id": 3515241,
      "authorName": "XuKong Ji",
      "votes": 0,
      "postDate": "2026-08-21T07:27:59.900000",
      "content": "<p>That's nice!!!</p>"
    },
    {
      "id": 3514885,
      "authorName": "Bull",
      "votes": 0,
      "postDate": "2026-08-20T10:48:42.737000",
      "content": "<blockquote>\n  <p><a href=\"https://www.kaggle.com/code/*******/*******\" target=\"_blank\">https://www.kaggle.com/code/*******/*******</a></p>\n  <p>Hey Claude/codex, extract any potentially useful insights from this public notebook.</p>\n</blockquote>\n<p>That’s all you need! You don’t even need to upvote a notebook just to fork it when you’re not sure whether it’ll be useful.</p>"
    },
    {
      "id": 3514963,
      "authorName": "Mattia Angeli",
      "votes": 1,
      "postDate": "2026-08-20T13:40:24.283000",
      "content": "<p>Maybe you could also ask Claude to scan the top scoring public notebooks and tell you which are original work, and which are a fork with one hyperparameter changed and a new name on it. The credits section is usually the part that got removed.</p>\n<p>Then you can decide if and who to upvote — and who to report.</p>"
    },
    {
      "id": 3515245,
      "authorName": "XuKong Ji",
      "votes": -1,
      "postDate": "2026-08-21T07:39:13.990000",
      "content": "<p>Seeming likes token is all i need😅</p>"
    }
  ],
  "index": {
    "id": "736268",
    "title": "Too many noise public notebookS.",
    "authorName": "",
    "commentCount": "13",
    "votes": "13",
    "postDate": "2026-08-20 04:31:34.063000"
  },
  "competition": "rsna-knee-abnormality-detection"
}