![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F812001%2Fc5199e3863de8103fb5a09ee06591df3%2Fdownload%20(9).png?generation=1745507168405558&alt=media\")
A few incorrect labels in the test dataset have been identified and updated. This requires a rescore of all submissions. The leaderboard will be in a moving state until it is completed. I'll update this thread once complete.
", "messages": [ { "id": 3185604, "postDate": "2025-04-23T15:10:03.027Z", "content": "A few incorrect labels in the test dataset have been identified and updated. This requires a rescore of all submissions. The leaderboard will be in a moving state until it is completed. I'll update this thread once complete.
", "rawMarkdown": "A few incorrect labels in the test dataset have been identified and updated. This requires a rescore of all submissions. The leaderboard will be in a moving state until it is completed. I'll update this thread once complete.", "votes": 25 }, { "id": 3185864, "postDate": "2025-04-23T23:44:57.637Z", "content": "0.553 to 0.785 …
", "rawMarkdown": "0.553 to 0.785 ...", "votes": 3, "replies": [ { "id": 3185883, "postDate": "2025-04-24T01:10:09.810Z", "content": "That is indeed crazy, wonder how many incorect labels if you have this much change in score.
", "rawMarkdown": "That is indeed crazy, wonder how many incorect labels if you have this much change in score." }, { "id": 3185954, "postDate": "2025-04-24T03:31:04.720Z", "content": "okay, lol, so it’s possible. i also had notes of a submission that scored 0.65 which is now at 0.79, thought it’s too much of a jump to be legit
\nL.E.: found screenshots, 0.652 to 0.790 confirmed as a jump. not at the 0.553 to 0.785 level, but definitely doesn't appear to be only 10 false negatives being corrected, looks like it's more.
", "rawMarkdown": "okay, lol, so it’s possible. i also had notes of a submission that scored 0.65 which is now at 0.79, thought it’s too much of a jump to be legit\n\nL.E.: found screenshots, 0.652 to 0.790 confirmed as a jump. not at the 0.553 to 0.785 level, but definitely doesn't appear to be only 10 false negatives being corrected, looks like it's more.", "votes": 1 }, { "id": 3185982, "postDate": "2025-04-24T04:33:41.583Z", "content": "That's insane. No way few corrected images would cause such a massive change in the score.
", "rawMarkdown": "That's insane. No way few corrected images would cause such a massive change in the score.", "votes": 1 } ] }, { "id": 3186813, "postDate": "2025-04-25T07:44:09.567Z", "content": "Am I the only one who lost like 70 places after the recalculation and had a post-recalculaiton score LOWERED? 🥹
", "rawMarkdown": "Am I the only one who lost like 70 places after the recalculation and had a post-recalculaiton score LOWERED? 🥹", "votes": 1, "replies": [ { "id": 3186845, "postDate": "2025-04-25T08:19:22.800Z", "content": "Previous best LB went down by 0.03 or so, but had another submission surpass previous best by 0.004 (it had an increase of 0.014)
", "rawMarkdown": "Previous best LB went down by 0.03 or so, but had another submission surpass previous best by 0.004 (it had an increase of 0.014)", "votes": 1, "replies": [ { "id": 3186852, "postDate": "2025-04-25T08:32:03.260Z", "content": "Same( 0.03 loss), probably the model found no flagella on an image where there was one.
", "rawMarkdown": "Same( 0.03 loss), probably the model found no flagella on an image where there was one.", "votes": 1 }, { "id": 3186860, "postDate": "2025-04-25T08:43:13.390Z", "content": "Exactly, and before, it had a 'boost' because it had better precision without having the recall affected. With the introduction of those new labels, recall went down.
\nA good trick is to identify the models which identify those motors and have them ready to ensemble with models with decent scores that could not identify those motors. This would be a great tool to maximize recall of motors.
\nThe leaderboard is super packed, I am really curious how this competition will turn out!
", "rawMarkdown": "Exactly, and before, it had a 'boost' because it had better precision without having the recall affected. With the introduction of those new labels, recall went down. \n\nA good trick is to identify the models which identify those motors and have them ready to ensemble with models with decent scores that could not identify those motors. This would be a great tool to maximize recall of motors.\n\nThe leaderboard is super packed, I am really curious how this competition will turn out!", "votes": 2 } ] } ] }, { "id": 3185860, "postDate": "2025-04-23T23:21:48.833Z", "content": "Man I though it has some bugs cause there's unreasonable result in my experiments. Now it all makes sense.
", "rawMarkdown": "Man I though it has some bugs cause there's unreasonable result in my experiments. Now it all makes sense.", "votes": 1 }, { "id": 3185786, "postDate": "2025-04-23T20:27:56.677Z", "content": "from 0.757 to 0.831 wow
", "rawMarkdown": "from 0.757 to 0.831 wow", "votes": 1, "replies": [ { "id": 3185788, "postDate": "2025-04-23T20:32:45.493Z", "content": "Definitely much better than the other direction! 😀
", "rawMarkdown": "Definitely much better than the other direction! 😀", "votes": 2 }, { "id": 3185820, "postDate": "2025-04-23T22:03:20.157Z", "content": "What on earth did you do? Haha I was hoping we were going to be in a similar boat, seems not hahaha
", "rawMarkdown": "What on earth did you do? Haha I was hoping we were going to be in a similar boat, seems not hahaha", "replies": [ { "id": 3185881, "postDate": "2025-04-24T01:07:56.777Z", "content": "To be honest I just trusted local metrics more than the public score. I tuned my solution based on local cv and it lead to this score. My experimental results make more sense now after this rescore.
", "rawMarkdown": "To be honest I just trusted local metrics more than the public score. I tuned my solution based on local cv and it lead to this score. My experimental results make more sense now after this rescore.", "votes": 2 } ] }, { "id": 3185879, "postDate": "2025-04-24T01:07:02.547Z", "content": "From 0.7 to 0.82, are you sure just a FEW incorrect labels?
", "rawMarkdown": "From 0.7 to 0.82, are you sure just a FEW incorrect labels?", "votes": 3, "replies": [ { "id": 3185882, "postDate": "2025-04-24T01:08:35.237Z", "content": "Hahaha! probably couple dozen
", "rawMarkdown": "Hahaha! probably couple dozen" } ] } ] }, { "id": 3186162, "postDate": "2025-04-24T09:34:32.923Z", "content": "Mistakes in the test set raise questions about the training set as well. Are there known issues in the training set that should maybe be addresses as well?
", "rawMarkdown": "Mistakes in the test set raise questions about the training set as well. Are there known issues in the training set that should maybe be addresses as well?", "votes": 2, "replies": [ { "id": 3186165, "postDate": "2025-04-24T09:40:35.327Z", "content": "I observe some issues in training set which are raised by @hengck23 in his previous post.
", "rawMarkdown": "I observe some issues in training set which are raised by @hengck23 in his previous post.", "votes": 1, "replies": [ { "id": 3186168, "postDate": "2025-04-24T09:44:57.983Z", "content": "I am making some new submission as my results flip flopped. Will close my thread and start a new one.
\nThis also shows how conclusion can change 180 when new information are given. Sometimes machine learning analysis are just story telling and hallucinations ( trying to make the most probable explanations which may be entirely false )
", "rawMarkdown": "I am making some new submission as my results flip flopped. Will close my thread and start a new one.\n\nThis also shows how conclusion can change 180 when new information are given. Sometimes machine learning analysis are just story telling and hallucinations ( trying to make the most probable explanations which may be entirely false )", "votes": 5 }, { "id": 3186173, "postDate": "2025-04-24T09:57:35.223Z", "content": "Yeah. I'm really enjoy discovering and pursuing the truth of data. That's why I put a lot of work in this comp.
", "rawMarkdown": "Yeah. I'm really enjoy discovering and pursuing the truth of data. That's why I put a lot of work in this comp.", "votes": 1 }, { "id": 3186174, "postDate": "2025-04-24T10:04:16.103Z", "content": "Same, a paradigm shift is required. I was very confused about some experiments, and after 200 LB submissions, trying to explain why some things were happening was hard. Some improvements were basically being punished and vice versa.
", "rawMarkdown": "Same, a paradigm shift is required. I was very confused about some experiments, and after 200 LB submissions, trying to explain why some things were happening was hard. Some improvements were basically being punished and vice versa.", "votes": 1 }, { "id": 3186324, "postDate": "2025-04-24T13:47:31.240Z", "content": "too bad that i did not record my old score. if not this is a good case study on how wrong labels affect cv/lb and how to detect wrong label, etc
", "rawMarkdown": "too bad that i did not record my old score. if not this is a good case study on how wrong labels affect cv/lb and how to detect wrong label, etc" }, { "id": 3186371, "postDate": "2025-04-24T15:07:03.147Z", "content": "I downloaded part of my score before update, the new score makes more sence than old one.
thanks for the graph. if you set color= threshold (or cv metric like recall, precision, etc), maybe you see something
", "rawMarkdown": "thanks for the graph. if you set color= threshold (or cv metric like recall, precision, etc), maybe you see something" }, { "id": 3186840, "postDate": "2025-04-25T08:09:01.423Z", "content": "I cannot seem to find the post of @hengck23 you mentioned @tom99763 . Could you please link it?
", "rawMarkdown": "I cannot seem to find the post of @hengck23 you mentioned @tom99763 . Could you please link it?" } ] } ] }, { "id": 3185933, "postDate": "2025-04-24T02:58:21.900Z", "content": "My ranking has probably improved by about 10 places.😁
", "rawMarkdown": "My ranking has probably improved by about 10 places.😁", "votes": 2 }, { "id": 3185614, "postDate": "2025-04-23T15:23:28.053Z", "content": "A few is roughly how many? Will we be able to see the difference on the LB?
", "rawMarkdown": "A few is roughly how many? Will we be able to see the difference on the LB?", "votes": 2, "replies": [ { "id": 3185619, "postDate": "2025-04-23T15:37:44.377Z", "content": "I hope about half of public.
", "rawMarkdown": "I hope about half of public.", "votes": 1 }, { "id": 3185674, "postDate": "2025-04-23T17:15:43.303Z", "content": "For the Public leaderboard, on the order of 10 corrections. These were almost all false negatives that now are correctly labeled with motors. There will definitely be changes on the LB, but by how much and in which direction will depend on how the model balances accuracy vs. recall.
", "rawMarkdown": "For the Public leaderboard, on the order of 10 corrections. These were almost all false negatives that now are correctly labeled with motors. There will definitely be changes on the LB, but by how much and in which direction will depend on how the model balances accuracy vs. recall.", "votes": 5 } ] }, { "id": 3185931, "postDate": "2025-04-24T02:56:58.020Z", "content": "Will the public scores of submitted notebooks under the submission tabs be changed too? Because for now, I see the change only in the leaderboard score.
", "rawMarkdown": "Will the public scores of submitted notebooks under the submission tabs be changed too? Because for now, I see the change only in the leaderboard score." }, { "id": 3185823, "postDate": "2025-04-23T22:03:54.243Z", "content": "i can see a correlation between my cv and lb
", "rawMarkdown": "i can see a correlation between my cv and lb" }, { "id": 3185648, "postDate": "2025-04-23T16:36:36.523Z", "content": "My previous model of LB 71.7 has turned into 80. That's an exaggeration. Is there something wrong?😅
", "rawMarkdown": "My previous model of LB 71.7 has turned into 80. That's an exaggeration. Is there something wrong?😅", "replies": [ { "id": 3185675, "postDate": "2025-04-23T17:17:09.837Z", "content": "A number of false negatives were corrected to have motors. It means your model was correctly predicting these, but you were wrongly penalized because the solution labels were incorrect.
", "rawMarkdown": "A number of false negatives were corrected to have motors. It means your model was correctly predicting these, but you were wrongly penalized because the solution labels were incorrect. ", "votes": 1, "replies": [ { "id": 3185677, "postDate": "2025-04-23T17:20:25.243Z", "content": "Thank you very much, is it possible to see how the scores change for each submission after that, as I don't remember the previous scores？
", "rawMarkdown": "Thank you very much, is it possible to see how the scores change for each submission after that, as I don't remember the previous scores？" }, { "id": 3185680, "postDate": "2025-04-23T17:26:12.183Z", "content": "Unfortunately, do don't have a way to display previous scores. Only the new scores will be listed under the Submissions tab.
", "rawMarkdown": "Unfortunately, do don't have a way to display previous scores. Only the new scores will be listed under the Submissions tab.", "votes": 1 } ] } ] }, { "id": 3185634, "postDate": "2025-04-23T15:58:59.270Z", "content": "Do we have any idea how long this will take to finish as a guess? As the CV score for this comp has not been immensely valuable the LB is fairly important to testing. Our team has nearly 200 tests we have run…
", "rawMarkdown": "Do we have any idea how long this will take to finish as a guess? As the CV score for this comp has not been immensely valuable the LB is fairly important to testing. Our team has nearly 200 tests we have run...", "replies": [ { "id": 3185670, "postDate": "2025-04-23T17:09:10.453Z", "content": "With the current rate of rescores, it should be completed in about 6 hours.
", "rawMarkdown": "With the current rate of rescores, it should be completed in about 6 hours.", "votes": 3 }, { "id": 3185875, "postDate": "2025-04-24T00:37:00.823Z", "content": "@cody11null Cody maybe next time we can team up together. I think you work really hard.
", "rawMarkdown": "@cody11null Cody maybe next time we can team up together. I think you work really hard.", "votes": 1, "replies": [ { "id": 3186024, "postDate": "2025-04-24T05:45:59.430Z", "content": "Sounds like a thrill, would love to learn more about the work you have been doing lately! 2 stellar GNNs in a row!
", "rawMarkdown": "Sounds like a thrill, would love to learn more about the work you have been doing lately! 2 stellar GNNs in a row!" } ] } ] }, { "id": 3185732, "postDate": "2025-04-23T18:26:37.713Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 3185864, "author_name": "I2nfinit3y", "author_url": "", "post_date": "2025-04-23T23:44:57.637000", "content": "0.553 to 0.785 …
", "votes": 3, "replies": [ { "id": 3185883, "author_name": "Yann Majewski", "author_url": "", "post_date": "2025-04-24T01:10:09.810000", "content": "That is indeed crazy, wonder how many incorect labels if you have this much change in score.
", "votes": 0, "replies": [] }, { "id": 3185954, "author_name": "Andrei Zamfir", "author_url": "", "post_date": "2025-04-24T03:31:04.720000", "content": "okay, lol, so it’s possible. i also had notes of a submission that scored 0.65 which is now at 0.79, thought it’s too much of a jump to be legit
\nL.E.: found screenshots, 0.652 to 0.790 confirmed as a jump. not at the 0.553 to 0.785 level, but definitely doesn't appear to be only 10 false negatives being corrected, looks like it's more.
", "votes": 1, "replies": [] }, { "id": 3185982, "author_name": "DennisSakva", "author_url": "", "post_date": "2025-04-24T04:33:41.583000", "content": "That's insane. No way few corrected images would cause such a massive change in the score.
", "votes": 1, "replies": [] } ] }, { "id": 3186813, "author_name": "DennisSakva", "author_url": "", "post_date": "2025-04-25T07:44:09.567000", "content": "Am I the only one who lost like 70 places after the recalculation and had a post-recalculaiton score LOWERED? 🥹
", "votes": 1, "replies": [ { "id": 3186845, "author_name": "Andrei Zamfir", "author_url": "", "post_date": "2025-04-25T08:19:22.800000", "content": "Previous best LB went down by 0.03 or so, but had another submission surpass previous best by 0.004 (it had an increase of 0.014)
", "votes": 1, "replies": [ { "id": 3186852, "author_name": "DennisSakva", "author_url": "", "post_date": "2025-04-25T08:32:03.260000", "content": "Same( 0.03 loss), probably the model found no flagella on an image where there was one.
", "votes": 1, "replies": [] }, { "id": 3186860, "author_name": "Andrei Zamfir", "author_url": "", "post_date": "2025-04-25T08:43:13.390000", "content": "Exactly, and before, it had a 'boost' because it had better precision without having the recall affected. With the introduction of those new labels, recall went down.
\nA good trick is to identify the models which identify those motors and have them ready to ensemble with models with decent scores that could not identify those motors. This would be a great tool to maximize recall of motors.
\nThe leaderboard is super packed, I am really curious how this competition will turn out!
", "votes": 2, "replies": [] } ] } ] }, { "id": 3185860, "author_name": "Tom", "author_url": "", "post_date": "2025-04-23T23:21:48.833000", "content": "Man I though it has some bugs cause there's unreasonable result in my experiments. Now it all makes sense.
", "votes": 1, "replies": [] }, { "id": 3185786, "author_name": "Yann Majewski", "author_url": "", "post_date": "2025-04-23T20:27:56.677000", "content": "from 0.757 to 0.831 wow
", "votes": 1, "replies": [ { "id": 3185788, "author_name": "inversion", "author_url": "", "post_date": "2025-04-23T20:32:45.493000", "content": "Definitely much better than the other direction! 😀
", "votes": 2, "replies": [] }, { "id": 3185820, "author_name": "Cody_Null", "author_url": "", "post_date": "2025-04-23T22:03:20.157000", "content": "What on earth did you do? Haha I was hoping we were going to be in a similar boat, seems not hahaha
", "votes": 0, "replies": [ { "id": 3185881, "author_name": "Yann Majewski", "author_url": "", "post_date": "2025-04-24T01:07:56.777000", "content": "To be honest I just trusted local metrics more than the public score. I tuned my solution based on local cv and it lead to this score. My experimental results make more sense now after this rescore.
", "votes": 2, "replies": [] } ] }, { "id": 3185879, "author_name": "sheep", "author_url": "", "post_date": "2025-04-24T01:07:02.547000", "content": "From 0.7 to 0.82, are you sure just a FEW incorrect labels?
", "votes": 3, "replies": [ { "id": 3185882, "author_name": "Yann Majewski", "author_url": "", "post_date": "2025-04-24T01:08:35.237000", "content": "Hahaha! probably couple dozen
", "votes": 0, "replies": [] } ] } ] }, { "id": 3186162, "author_name": "FabianIsensee", "author_url": "", "post_date": "2025-04-24T09:34:32.923000", "content": "Mistakes in the test set raise questions about the training set as well. Are there known issues in the training set that should maybe be addresses as well?
", "votes": 2, "replies": [ { "id": 3186165, "author_name": "Tom", "author_url": "", "post_date": "2025-04-24T09:40:35.327000", "content": "I observe some issues in training set which are raised by @hengck23 in his previous post.
", "votes": 1, "replies": [ { "id": 3186168, "author_name": "hengck23", "author_url": "", "post_date": "2025-04-24T09:44:57.983000", "content": "I am making some new submission as my results flip flopped. Will close my thread and start a new one.
\nThis also shows how conclusion can change 180 when new information are given. Sometimes machine learning analysis are just story telling and hallucinations ( trying to make the most probable explanations which may be entirely false )
", "votes": 5, "replies": [] }, { "id": 3186173, "author_name": "Tom", "author_url": "", "post_date": "2025-04-24T09:57:35.223000", "content": "Yeah. I'm really enjoy discovering and pursuing the truth of data. That's why I put a lot of work in this comp.
", "votes": 1, "replies": [] }, { "id": 3186174, "author_name": "Andrei Zamfir", "author_url": "", "post_date": "2025-04-24T10:04:16.103000", "content": "Same, a paradigm shift is required. I was very confused about some experiments, and after 200 LB submissions, trying to explain why some things were happening was hard. Some improvements were basically being punished and vice versa.
", "votes": 1, "replies": [] }, { "id": 3186324, "author_name": "hengck23", "author_url": "", "post_date": "2025-04-24T13:47:31.240000", "content": "too bad that i did not record my old score. if not this is a good case study on how wrong labels affect cv/lb and how to detect wrong label, etc
", "votes": 0, "replies": [] }, { "id": 3186371, "author_name": "sheep", "author_url": "", "post_date": "2025-04-24T15:07:03.147000", "content": "I downloaded part of my score before update, the new score makes more sence than old one.
thanks for the graph. if you set color= threshold (or cv metric like recall, precision, etc), maybe you see something
", "votes": 0, "replies": [] }, { "id": 3186840, "author_name": "FabianIsensee", "author_url": "", "post_date": "2025-04-25T08:09:01.423000", "content": "I cannot seem to find the post of @hengck23 you mentioned @tom99763 . Could you please link it?
", "votes": 0, "replies": [] } ] } ] }, { "id": 3185933, "author_name": "Less", "author_url": "", "post_date": "2025-04-24T02:58:21.900000", "content": "My ranking has probably improved by about 10 places.😁
", "votes": 2, "replies": [] }, { "id": 3185614, "author_name": "DennisSakva", "author_url": "", "post_date": "2025-04-23T15:23:28.053000", "content": "A few is roughly how many? Will we be able to see the difference on the LB?
", "votes": 2, "replies": [ { "id": 3185619, "author_name": "Ángel Jacinto Sánchez Ruiz", "author_url": "", "post_date": "2025-04-23T15:37:44.377000", "content": "I hope about half of public.
", "votes": 1, "replies": [] }, { "id": 3185674, "author_name": "inversion", "author_url": "", "post_date": "2025-04-23T17:15:43.303000", "content": "For the Public leaderboard, on the order of 10 corrections. These were almost all false negatives that now are correctly labeled with motors. There will definitely be changes on the LB, but by how much and in which direction will depend on how the model balances accuracy vs. recall.
", "votes": 5, "replies": [] } ] }, { "id": 3185931, "author_name": "Udit Jain", "author_url": "", "post_date": "2025-04-24T02:56:58.020000", "content": "Will the public scores of submitted notebooks under the submission tabs be changed too? Because for now, I see the change only in the leaderboard score.
", "votes": 0, "replies": [] }, { "id": 3185823, "author_name": "MengYe", "author_url": "", "post_date": "2025-04-23T22:03:54.243000", "content": "i can see a correlation between my cv and lb
", "votes": 0, "replies": [] }, { "id": 3185648, "author_name": "yyyy0201", "author_url": "", "post_date": "2025-04-23T16:36:36.523000", "content": "My previous model of LB 71.7 has turned into 80. That's an exaggeration. Is there something wrong?😅
", "votes": 0, "replies": [ { "id": 3185675, "author_name": "inversion", "author_url": "", "post_date": "2025-04-23T17:17:09.837000", "content": "A number of false negatives were corrected to have motors. It means your model was correctly predicting these, but you were wrongly penalized because the solution labels were incorrect.
", "votes": 1, "replies": [ { "id": 3185677, "author_name": "yyyy0201", "author_url": "", "post_date": "2025-04-23T17:20:25.243000", "content": "Thank you very much, is it possible to see how the scores change for each submission after that, as I don't remember the previous scores？
", "votes": 0, "replies": [] }, { "id": 3185680, "author_name": "inversion", "author_url": "", "post_date": "2025-04-23T17:26:12.183000", "content": "Unfortunately, do don't have a way to display previous scores. Only the new scores will be listed under the Submissions tab.
", "votes": 1, "replies": [] } ] } ] }, { "id": 3185634, "author_name": "Cody_Null", "author_url": "", "post_date": "2025-04-23T15:58:59.270000", "content": "Do we have any idea how long this will take to finish as a guess? As the CV score for this comp has not been immensely valuable the LB is fairly important to testing. Our team has nearly 200 tests we have run…
", "votes": 0, "replies": [ { "id": 3185670, "author_name": "inversion", "author_url": "", "post_date": "2025-04-23T17:09:10.453000", "content": "With the current rate of rescores, it should be completed in about 6 hours.
", "votes": 3, "replies": [] }, { "id": 3185875, "author_name": "Tom", "author_url": "", "post_date": "2025-04-24T00:37:00.823000", "content": "@cody11null Cody maybe next time we can team up together. I think you work really hard.
", "votes": 1, "replies": [ { "id": 3186024, "author_name": "Cody_Null", "author_url": "", "post_date": "2025-04-24T05:45:59.430000", "content": "Sounds like a thrill, would love to learn more about the work you have been doing lately! 2 stellar GNNs in a row!
", "votes": 0, "replies": [] } ] } ] }, { "id": 3185732, "author_name": "", "author_url": "", "post_date": "2025-04-23T18:26:37.713000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "3185604": "A few incorrect labels in the test dataset have been identified and updated. This requires a rescore of all submissions. The leaderboard will be in a moving state until it is completed. I'll update this thread once complete.", "3185864": "0.553 to 0.785 ...", "3186813": "Am I the only one who lost like 70 places after the recalculation and had a post-recalculaiton score LOWERED? 🥹", "3185860": "Man I though it has some bugs cause there's unreasonable result in my experiments. Now it all makes sense.", "3185786": "from 0.757 to 0.831 wow", "3186162": "Mistakes in the test set raise questions about the training set as well. Are there known issues in the training set that should maybe be addresses as well?", "3185933": "My ranking has probably improved by about 10 places.😁", "3185614": "A few is roughly how many? Will we be able to see the difference on the LB?", "3185931": "Will the public scores of submitted notebooks under the submission tabs be changed too? Because for now, I see the change only in the leaderboard score.", "3185823": "i can see a correlation between my cv and lb", "3185648": "My previous model of LB 71.7 has turned into 80. That's an exaggeration. Is there something wrong?😅", "3185634": "Do we have any idea how long this will take to finish as a guess? As the CV score for this comp has not been immensely valuable the LB is fairly important to testing. Our team has nearly 200 tests we have run...", "3185732": "" } }