Finally, someone achieved 0.80 LB. Congratulations @marcoledda29
", "messages": [ { "id": "3395187", "postDate": "01/22/2026 12:47:41", "content": "Finally, someone achieved 0.80 LB. Congratulations @marcoledda29
", "rawMarkdown": "Finally, someone achieved 0.80 LB. Congratulations @marcoledda29", "votes": null }, { "id": "3395200", "postDate": "01/22/2026 13:12:35", "content": "Normally, RMSE is quite hard to improve, so it’s really surprising that we were able to improve the weighted R² from 0.73 to 0.8 🥳 🥳 🥳
", "rawMarkdown": "Normally, RMSE is quite hard to improve, so it’s really surprising that we were able to improve the weighted R² from 0.73 to 0.8 🥳 🥳 🥳", "votes": null }, { "id": "3395218", "postDate": "01/22/2026 13:40:47", "content": "Thanks! It's a nice result, but it's \"just\" public leaderboard. Let's see what happens at the end! Good luck for the final week of the competition!
", "rawMarkdown": "Thanks! It's a nice result, but it's \"just\" public leaderboard. Let's see what happens at the end! Good luck for the final week of the competition!", "votes": null }, { "id": "3395429", "postDate": "01/23/2026 00:51:23", "content": "I don’t think that public and private LB have a big difference. I assume that shake up won’t be. Have a good luck!
", "rawMarkdown": "I don’t think that public and private LB have a big difference. I assume that shake up won’t be. Have a good luck!", "votes": null }, { "id": "3395430", "postDate": "01/23/2026 00:54:05", "content": "Haha yes, it’s pretty impressive
", "rawMarkdown": "Haha yes, it’s pretty impressive", "votes": null }, { "id": "3395431", "postDate": "01/23/2026 00:55:04", "content": "How are you so confident about that? :)
", "rawMarkdown": "How are you so confident about that? :)", "votes": null }, { "id": "3395434", "postDate": "01/23/2026 01:03:01", "content": "I think that all top participants have a stable CV, I’m not sure in 100% but the biggest part of me thinks that it’s true
", "rawMarkdown": "I think that all top participants have a stable CV, I’m not sure in 100% but the biggest part of me thinks that it’s true", "votes": null }, { "id": "3395436", "postDate": "01/23/2026 01:08:49", "content": "And I also wanna hear your opinion about shake up. What do u think?
", "rawMarkdown": "And I also wanna hear your opinion about shake up. What do u think?", "votes": null }, { "id": "3398085", "postDate": "01/28/2026 15:24:42", "content": "0.79 is understandable, but 0.81 is truly amazing. I think the top 1 solution must be completely different from the others.
", "rawMarkdown": "0.79 is understandable, but 0.81 is truly amazing. I think the top 1 solution must be completely different from the others.", "votes": null }, { "id": "3398145", "postDate": "01/28/2026 17:35:33", "content": "Shakeup is inevitable. Cross validation did not seem to work at all. People have reported CV of 0.78, 0.83 without data leakage but still score less than 0.70. So how is CV stable?
", "rawMarkdown": "Shakeup is inevitable. Cross validation did not seem to work at all. People have reported CV of 0.78, 0.83 without data leakage but still score less than 0.70. So how is CV stable?", "votes": null }, { "id": "3398148", "postDate": "01/28/2026 17:38:04", "content": "Only luck can save participants. As suspected by some of us, data splits must be based on years. So if the private test data is from years 2017, 2018, …, then we may have very different biomass targets.
", "rawMarkdown": "Only luck can save participants. As suspected by some of us, data splits must be based on years. So if the private test data is from years 2017, 2018, ..., then we may have very different biomass targets.", "votes": null }, { "id": "3398280", "postDate": "01/28/2026 23:58:25", "content": "No, your CV is actually working but there is another issue. Train and test datas has a very big difference and for this reason people who achieved score 0.74 have been found a way to solve this
", "rawMarkdown": "No, your CV is actually working but there is another issue. Train and test datas has a very big difference and for this reason people who achieved score 0.74 have been found a way to solve this", "votes": null }, { "id": "3398282", "postDate": "01/29/2026 00:03:47", "content": "Congratulations top-2
", "rawMarkdown": "Congratulations top-2", "votes": null }, { "id": "3398283", "postDate": "01/29/2026 00:06:16", "content": "😂😂🤣🤣😭😭
", "rawMarkdown": "😂😂🤣🤣😭😭", "votes": null }, { "id": "3398285", "postDate": "01/29/2026 00:06:43", "content": "Thank you. Surprisingly, post-processing actually helps.
", "rawMarkdown": "Thank you. Surprisingly, post-processing actually helps.", "votes": null }, { "id": "3398289", "postDate": "01/29/2026 00:17:49", "content": "The shake-up is smaller than I thought it would be… I was honestly expecting something much bigger. This task is inherently pretty random, so there’s really no accuracy to speak of. There’s no way the leaderboard can stay stable.
", "rawMarkdown": "The shake-up is smaller than I thought it would be... I was honestly expecting something much bigger. This task is inherently pretty random, so there’s really no accuracy to speak of. There’s no way the leaderboard can stay stable.", "votes": null }, { "id": "3398301", "postDate": "01/29/2026 00:31:06", "content": "as someone who dropped around 800 ranks while choosing relatively conservative submissions, I think the shake up was enough for us
", "rawMarkdown": "as someone who dropped around 800 ranks while choosing relatively conservative submissions, I think the shake up was enough for us", "votes": null }, { "id": "3398307", "postDate": "01/29/2026 00:36:17", "content": "Sorry to hear that. One thing I noticed is that ensembling never really worked for this task, which is quite strange. A single model can easily outperform dozens of carefully selected models combined, which again highlights how random this task is.
", "rawMarkdown": "Sorry to hear that. One thing I noticed is that ensembling never really worked for this task, which is quite strange. A single model can easily outperform dozens of carefully selected models combined, which again highlights how random this task is.", "votes": null } ], "comments": [ { "id": 3395200, "author_name": "quan0095", "author_url": "", "post_date": "01/22/2026 13:12:35", "content": "Normally, RMSE is quite hard to improve, so it’s really surprising that we were able to improve the weighted R² from 0.73 to 0.8 🥳 🥳 🥳
", "votes": null, "replies": [ { "id": 3395430, "author_name": "artemgoncarov", "author_url": "", "post_date": "01/23/2026 00:54:05", "content": "Haha yes, it’s pretty impressive
", "votes": null, "replies": [] } ] }, { "id": 3395218, "author_name": "marcoledda29", "author_url": "", "post_date": "01/22/2026 13:40:47", "content": "Thanks! It's a nice result, but it's \"just\" public leaderboard. Let's see what happens at the end! Good luck for the final week of the competition!
", "votes": null, "replies": [ { "id": 3395429, "author_name": "artemgoncarov", "author_url": "", "post_date": "01/23/2026 00:51:23", "content": "I don’t think that public and private LB have a big difference. I assume that shake up won’t be. Have a good luck!
", "votes": null, "replies": [ { "id": 3395431, "author_name": "nischaydnk", "author_url": "", "post_date": "01/23/2026 00:55:04", "content": "How are you so confident about that? :)
", "votes": null, "replies": [ { "id": 3395434, "author_name": "artemgoncarov", "author_url": "", "post_date": "01/23/2026 01:03:01", "content": "I think that all top participants have a stable CV, I’m not sure in 100% but the biggest part of me thinks that it’s true
", "votes": null, "replies": [] }, { "id": 3395436, "author_name": "artemgoncarov", "author_url": "", "post_date": "01/23/2026 01:08:49", "content": "And I also wanna hear your opinion about shake up. What do u think?
", "votes": null, "replies": [ { "id": 3398145, "author_name": "jeannkouagou", "author_url": "", "post_date": "01/28/2026 17:35:33", "content": "Shakeup is inevitable. Cross validation did not seem to work at all. People have reported CV of 0.78, 0.83 without data leakage but still score less than 0.70. So how is CV stable?
", "votes": null, "replies": [ { "id": 3398148, "author_name": "jeannkouagou", "author_url": "", "post_date": "01/28/2026 17:38:04", "content": "Only luck can save participants. As suspected by some of us, data splits must be based on years. So if the private test data is from years 2017, 2018, …, then we may have very different biomass targets.
", "votes": null, "replies": [] }, { "id": 3398280, "author_name": "artemgoncarov", "author_url": "", "post_date": "01/28/2026 23:58:25", "content": "No, your CV is actually working but there is another issue. Train and test datas has a very big difference and for this reason people who achieved score 0.74 have been found a way to solve this
", "votes": null, "replies": [] } ] } ] } ] }, { "id": 3398289, "author_name": "yannan90", "author_url": "", "post_date": "01/29/2026 00:17:49", "content": "The shake-up is smaller than I thought it would be… I was honestly expecting something much bigger. This task is inherently pretty random, so there’s really no accuracy to speak of. There’s no way the leaderboard can stay stable.
", "votes": null, "replies": [ { "id": 3398301, "author_name": "tahaalshatiri", "author_url": "", "post_date": "01/29/2026 00:31:06", "content": "as someone who dropped around 800 ranks while choosing relatively conservative submissions, I think the shake up was enough for us
", "votes": null, "replies": [ { "id": 3398307, "author_name": "yannan90", "author_url": "", "post_date": "01/29/2026 00:36:17", "content": "Sorry to hear that. One thing I noticed is that ensembling never really worked for this task, which is quite strange. A single model can easily outperform dozens of carefully selected models combined, which again highlights how random this task is.
", "votes": null, "replies": [] } ] } ] } ] } ] }, { "id": 3398085, "author_name": "quan0095", "author_url": "", "post_date": "01/28/2026 15:24:42", "content": "0.79 is understandable, but 0.81 is truly amazing. I think the top 1 solution must be completely different from the others.
", "votes": null, "replies": [ { "id": 3398282, "author_name": "kongpasom", "author_url": "", "post_date": "01/29/2026 00:03:47", "content": "Congratulations top-2
", "votes": null, "replies": [ { "id": 3398285, "author_name": "quan0095", "author_url": "", "post_date": "01/29/2026 00:06:43", "content": "Thank you. Surprisingly, post-processing actually helps.
", "votes": null, "replies": [] } ] } ] }, { "id": 3398283, "author_name": "playwithme", "author_url": "", "post_date": "01/29/2026 00:06:16", "content": "😂😂🤣🤣😭😭
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3395187": "Finally, someone achieved 0.80 LB. Congratulations @marcoledda29", "3395200": "Normally, RMSE is quite hard to improve, so it’s really surprising that we were able to improve the weighted R² from 0.73 to 0.8 🥳 🥳 🥳", "3395218": "Thanks! It's a nice result, but it's \"just\" public leaderboard. Let's see what happens at the end! Good luck for the final week of the competition!", "3395429": "I don’t think that public and private LB have a big difference. I assume that shake up won’t be. Have a good luck!", "3395430": "Haha yes, it’s pretty impressive", "3395431": "How are you so confident about that? :)", "3395434": "I think that all top participants have a stable CV, I’m not sure in 100% but the biggest part of me thinks that it’s true", "3395436": "And I also wanna hear your opinion about shake up. What do u think?", "3398085": "0.79 is understandable, but 0.81 is truly amazing. I think the top 1 solution must be completely different from the others.", "3398145": "Shakeup is inevitable. Cross validation did not seem to work at all. People have reported CV of 0.78, 0.83 without data leakage but still score less than 0.70. So how is CV stable?", "3398148": "Only luck can save participants. As suspected by some of us, data splits must be based on years. So if the private test data is from years 2017, 2018, ..., then we may have very different biomass targets.", "3398280": "No, your CV is actually working but there is another issue. Train and test datas has a very big difference and for this reason people who achieved score 0.74 have been found a way to solve this", "3398282": "Congratulations top-2", "3398283": "😂😂🤣🤣😭😭", "3398285": "Thank you. Surprisingly, post-processing actually helps.", "3398289": "The shake-up is smaller than I thought it would be... I was honestly expecting something much bigger. This task is inherently pretty random, so there’s really no accuracy to speak of. There’s no way the leaderboard can stay stable.", "3398301": "as someone who dropped around 800 ranks while choosing relatively conservative submissions, I think the shake up was enough for us", "3398307": "Sorry to hear that. One thing I noticed is that ensembling never really worked for this task, which is quite strange. A single model can easily outperform dozens of carefully selected models combined, which again highlights how random this task is." }, "source": "meta" }