![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8722753%2Fe9bf03e6fd1c12b02e2552bc4c923d20%2FSin%20ttulo.jpg?generation=1757372591604823&alt=media\")
Hi, I'm new to this competition. Reading through discussions, I noticed that most people on the LB has 0.69, which is equal to the performance of the public notebook 0.69 LB notebook.
\nHowever, it was noted in multiple discussions (e.g. here) that this notebook scores 0.5 now after some scoring / test set updates. In another thread, the organizers told they won't re-score submissions
\nDoes that mean the Public LB is useless now? Does it has a mix of old and new scores?
", "messages": [ { "id": 3287120, "postDate": "2025-09-10T22:19:53.847Z", "content": "No, the public leaderboard is still valid. Throughout the challenge we have made only very small changes to the test set including removal of a <5% of cases that had unfixable issues (and were probably not being accurately predicted anyway). We contemplated wiping the public leaderboard, but decided that it was not necessary since the dataset changes were so small. I have not heard that the public notebook had such a significant change in score, but if that is true then I suspect its for technical reasons.
", "rawMarkdown": "No, the public leaderboard is still valid. Throughout the challenge we have made only very small changes to the test set including removal of a <5% of cases that had unfixable issues (and were probably not being accurately predicted anyway). We contemplated wiping the public leaderboard, but decided that it was not necessary since the dataset changes were so small. I have not heard that the public notebook had such a significant change in score, but if that is true then I suspect its for technical reasons.", "votes": 1 }, { "id": 3285945, "postDate": "2025-09-09T00:11:25.353Z", "content": "No, that is not the case. There was a timeframe where the dataset had some issues and notebooks scored 0.5, but those submissions have been rerun and properly scored, so the public LB is all the correct scores now.
", "rawMarkdown": "No, that is not the case. There was a timeframe where the dataset had some issues and notebooks scored 0.5, but those submissions have been rerun and properly scored, so the public LB is all the correct scores now.", "votes": 1 }, { "id": 3285932, "postDate": "2025-09-08T23:03:13.853Z", "content": "Hi. I think you are talking about NeurIPS competition. That one remains stable after dataset minor fixes.
\n
I've read that discussion. No idea. I'll try to rerun it.
", "rawMarkdown": "Hi. I think you are talking about NeurIPS competition. That one remains stable after dataset minor fixes.\n\n\n\nI've read that discussion. No idea. I'll try to rerun it.", "votes": 1 }, { "id": 3285919, "postDate": "2025-09-08T22:25:55.110Z", "content": "Hi, I'm new to this competition. Reading through discussions, I noticed that most people on the LB has 0.69, which is equal to the performance of the public notebook 0.69 LB notebook.
\nHowever, it was noted in multiple discussions (e.g. here) that this notebook scores 0.5 now after some scoring / test set updates. In another thread, the organizers told they won't re-score submissions
\nDoes that mean the Public LB is useless now? Does it has a mix of old and new scores?
", "rawMarkdown": "Hi, I'm new to this competition. Reading through discussions, I noticed that most people on the LB has 0.69, which is equal to the performance of the public notebook [0.69 LB notebook](https://www.kaggle.com/code/yosukeyama/rsna2025-32ch-img-infer-lb-0-69-share).\n\nHowever, it was noted in multiple discussions [(e.g. here)](https://www.kaggle.com/code/yosukeyama/rsna2025-32ch-img-infer-lb-0-69-share/comments) that this notebook scores 0.5 now after some scoring / test set updates. In [another thread](https://www.kaggle.com/competitions/rsna-intracranial-aneurysm-detection/discussion/600908), the organizers told they won't re-score submissions\n\n**Does that mean the Public LB is useless now? Does it has a mix of old and new scores?**", "votes": 1 } ], "comments": [ { "id": 3287120, "author_name": "Evan Calabrese", "author_url": "", "post_date": "2025-09-10T22:19:53.847000", "content": "No, the public leaderboard is still valid. Throughout the challenge we have made only very small changes to the test set including removal of a <5% of cases that had unfixable issues (and were probably not being accurately predicted anyway). We contemplated wiping the public leaderboard, but decided that it was not necessary since the dataset changes were so small. I have not heard that the public notebook had such a significant change in score, but if that is true then I suspect its for technical reasons.
", "votes": 1, "replies": [] }, { "id": 3285945, "author_name": "Satwik", "author_url": "", "post_date": "2025-09-09T00:11:25.353000", "content": "No, that is not the case. There was a timeframe where the dataset had some issues and notebooks scored 0.5, but those submissions have been rerun and properly scored, so the public LB is all the correct scores now.
", "votes": 1, "replies": [] }, { "id": 3285932, "author_name": "Ángel Jacinto Sánchez Ruiz", "author_url": "", "post_date": "2025-09-08T23:03:13.853000", "content": "Hi. I think you are talking about NeurIPS competition. That one remains stable after dataset minor fixes.
\n
I've read that discussion. No idea. I'll try to rerun it.
", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "3287120": "No, the public leaderboard is still valid. Throughout the challenge we have made only very small changes to the test set including removal of a <5% of cases that had unfixable issues (and were probably not being accurately predicted anyway). We contemplated wiping the public leaderboard, but decided that it was not necessary since the dataset changes were so small. I have not heard that the public notebook had such a significant change in score, but if that is true then I suspect its for technical reasons.", "3285945": "No, that is not the case. There was a timeframe where the dataset had some issues and notebooks scored 0.5, but those submissions have been rerun and properly scored, so the public LB is all the correct scores now.", "3285932": "Hi. I think you are talking about NeurIPS competition. That one remains stable after dataset minor fixes.\n\n\n\nI've read that discussion. No idea. I'll try to rerun it.", "3285919": "Hi, I'm new to this competition. Reading through discussions, I noticed that most people on the LB has 0.69, which is equal to the performance of the public notebook [0.69 LB notebook](https://www.kaggle.com/code/yosukeyama/rsna2025-32ch-img-infer-lb-0-69-share).\n\nHowever, it was noted in multiple discussions [(e.g. here)](https://www.kaggle.com/code/yosukeyama/rsna2025-32ch-img-infer-lb-0-69-share/comments) that this notebook scores 0.5 now after some scoring / test set updates. In [another thread](https://www.kaggle.com/competitions/rsna-intracranial-aneurysm-detection/discussion/600908), the organizers told they won't re-score submissions\n\n**Does that mean the Public LB is useless now? Does it has a mix of old and new scores?**" } }