{
  "id": 611272,
  "title": "Dataset Clarification",
  "url": "/competitions/alpha-radar-solana-sprint/discussion/611272",
  "author_name": "",
  "post_date": "2025-10-09T23:05:28.029034900Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>In the dataset, I can see the big data and its corresponding target tokens, so is this what we use to train? There is also the sample with target tokens, if this is for testing, then why the target tokens are also provided? Pls clarify the dataset so that we can know what to do. Thank you.</p>",
  "messages": [
    {
      "id": "3300233",
      "postDate": "10/09/2025 23:05:28",
      "content": "<p>In the dataset, I can see the big data and its corresponding target tokens, so is this what we use to train? There is also the sample with target tokens, if this is for testing, then why the target tokens are also provided? Pls clarify the dataset so that we can know what to do. Thank you.</p>",
      "rawMarkdown": "In the dataset, I can see the big data and its corresponding target tokens, so is this what we use to train? There is also the sample with target tokens, if this is for testing, then why the target tokens are also provided? Pls clarify the dataset so that we can know what to do. Thank you.",
      "votes": null
    },
    {
      "id": "3300234",
      "postDate": "10/09/2025 23:12:13",
      "content": "<p>The big dataset and target tokens are the ones to use. The sample and sample target are just one of the proportional chunks from the larger set. It was a smaller upload for the main page, that's all. </p>",
      "rawMarkdown": "The big dataset and target tokens are the ones to use. The sample and sample target are just one of the proportional chunks from the larger set. It was a smaller upload for the main page, that's all.",
      "votes": null
    },
    {
      "id": "3300248",
      "postDate": "10/10/2025 00:53:23",
      "content": "<p>So is this big dataset used for both training and evaluation? I can see that many people get 1 by taking the given target tokens for evaluation. Then the leaderboard is no longer meaningful right?</p>",
      "rawMarkdown": "So is this big dataset used for both training and evaluation? I can see that many people get 1 by taking the given target tokens for evaluation. Then the leaderboard is no longer meaningful right?",
      "votes": null
    },
    {
      "id": "3300251",
      "postDate": "10/10/2025 01:03:05",
      "content": "<p>Yea, getting a 1 perfect result that way isn't going to be considered. The goal is to build a model that predicts for the target tokens based on their features, not just one that identifies tokens from an existing obvious subset. It's the overlap that will be rewarded. </p>",
      "rawMarkdown": "Yea, getting a 1 perfect result that way isn't going to be considered. The goal is to build a model that predicts for the target tokens based on their features, not just one that identifies tokens from an existing obvious subset. It's the overlap that will be rewarded.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3300234,
      "author_name": "twainayar",
      "author_url": "",
      "post_date": "10/09/2025 23:12:13",
      "content": "<p>The big dataset and target tokens are the ones to use. The sample and sample target are just one of the proportional chunks from the larger set. It was a smaller upload for the main page, that's all. </p>",
      "votes": null,
      "replies": [
        {
          "id": 3300248,
          "author_name": "lctngdng",
          "author_url": "",
          "post_date": "10/10/2025 00:53:23",
          "content": "<p>So is this big dataset used for both training and evaluation? I can see that many people get 1 by taking the given target tokens for evaluation. Then the leaderboard is no longer meaningful right?</p>",
          "votes": null,
          "replies": [
            {
              "id": 3300251,
              "author_name": "twainayar",
              "author_url": "",
              "post_date": "10/10/2025 01:03:05",
              "content": "<p>Yea, getting a 1 perfect result that way isn't going to be considered. The goal is to build a model that predicts for the target tokens based on their features, not just one that identifies tokens from an existing obvious subset. It's the overlap that will be rewarded. </p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3300233": "In the dataset, I can see the big data and its corresponding target tokens, so is this what we use to train? There is also the sample with target tokens, if this is for testing, then why the target tokens are also provided? Pls clarify the dataset so that we can know what to do. Thank you.",
    "3300234": "The big dataset and target tokens are the ones to use. The sample and sample target are just one of the proportional chunks from the larger set. It was a smaller upload for the main page, that's all.",
    "3300248": "So is this big dataset used for both training and evaluation? I can see that many people get 1 by taking the given target tokens for evaluation. Then the leaderboard is no longer meaningful right?",
    "3300251": "Yea, getting a 1 perfect result that way isn't going to be considered. The goal is to build a model that predicts for the target tokens based on their features, not just one that identifies tokens from an existing obvious subset. It's the overlap that will be rewarded."
  },
  "source": "meta"
}