{ "id": 585128, "title": "What's your guess of the number of teams that will break the 15.0 barrier?", "url": "/competitions/waveform-inversion/discussion/585128", "author_name": "c-number", "post_date": "2025-06-18T06:45:40.923000", "votes": 15, "comment_count": 39, "views": 0, "content": "

My guess is 3 teams.

", "messages": [ { "id": 3226841, "postDate": "2025-06-18T06:45:40.923Z", "content": "

My guess is 3 teams.

", "rawMarkdown": "My guess is 3 teams.", "votes": 15 }, { "id": 3232029, "postDate": "2025-06-25T09:44:46.667Z", "content": "

the speed of progress of the leaderboard in the last few days is still shocking to me, even though i expect as dramatic improvement. Based on my understanding of the \"solution\", my conclusion is that many top teams have many very good GPUs. I suppose they are in the range of A100,H100 (maybe some have H200, any have B200?),4090 and 5090. Hard to imagine top team proably have at least [masked] cards ? …

\n

the scale of computation resource today is different from those when i first joined kaggle more than 10 years ago .

\n

\"\"
\nGPU working hard at last days of competition … go go go!

", "rawMarkdown": "the speed of progress of the leaderboard in the last few days is still shocking to me, even though i expect as dramatic improvement. Based on my understanding of the \"solution\", my conclusion is that many top teams have many very good GPUs. I suppose they are in the range of A100,H100 (maybe some have H200, any have B200?),4090 and 5090. Hard to imagine top team proably have at least [masked] cards ? ...\n\nthe scale of computation resource today is different from those when i first joined kaggle more than 10 years ago .\n\n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Feac8dfd9ab51ba3d2424f3540d9e005d%2FSelection_999(8272).png?generation=1750846093211093&alt=media)\nGPU working hard at last days of competition ... go go go!", "votes": 4, "replies": [ { "id": 3232074, "postDate": "2025-06-25T10:41:59.680Z", "content": "

Someone said he had spare GB200.

", "rawMarkdown": "Someone said he had spare GB200.", "replies": [ { "id": 3233096, "postDate": "2025-06-26T13:28:29.383Z", "content": "

lol that's me
\nif only I had these a few months ago in AIMO-2 :/

", "rawMarkdown": "lol that's me \nif only I had these a few months ago in AIMO-2 :/", "votes": 1 } ] }, { "id": 3232077, "postDate": "2025-06-25T10:44:49.413Z", "content": "

You say \"solution\" as if it's not a real solution.
\nEven if you know what to do, actually doing it is still not trivial! 🤣

", "rawMarkdown": "You say \"solution\" as if it's not a real solution.\nEven if you know what to do, actually doing it is still not trivial! 🤣", "votes": 1, "replies": [ { "id": 3232081, "postDate": "2025-06-25T10:49:23.367Z", "content": "

of course! the details are in the devil!

\n

everyone knows what to do,
\nbut only one would survives

\n
\n

last time when i was in company, my project manager boss would suggest \"if detection is not good enough, then do tracking to improve the accuracy\"

\n

then in my mind, i would grumble \"you have to tell me how to implement the tracking!!! tracking is not the solution. the details are\"

\n
\n

but then when you explain your solution to your client, you don't want to confuse them with the details and also not to spill out the secret sauce, you would say \" the winning part of our solution when compared to competitors in the marked is that we added tracking to normal detection\".

", "rawMarkdown": "of course! the details are in the devil!\n\neveryone knows what to do,\nbut only one would survives \n\n----\n\nlast time when i was in company, my project manager boss would suggest \"if detection is not good enough, then do tracking to improve the accuracy\"\n\nthen in my mind, i would grumble \"you have to tell me how to implement the tracking!!! tracking is not the solution. the details are\"\n\n---\n\nbut then when you explain your solution to your client, you don't want to confuse them with the details and also not to spill out the secret sauce, you would say \" the winning part of our solution when compared to competitors in the marked is that we added tracking to normal detection\".", "votes": 1 }, { "id": 3232186, "postDate": "2025-06-25T13:23:41.500Z", "content": "
\n

everyone knows what to do,

\n
\n

I doubt it. I don't think I do to start with.

", "rawMarkdown": "> everyone knows what to do,\n\nI doubt it. I don't think I do to start with." }, { "id": 3232402, "postDate": "2025-06-25T16:50:03.953Z", "content": "

Most teams are still calculating—I believe it's just a matter of time before the rest are done.

", "rawMarkdown": "Most teams are still calculating—I believe it's just a matter of time before the rest are done." } ] } ] }, { "id": 3227217, "postDate": "2025-06-18T17:01:12.513Z", "content": "

I am more optimistic. I even think near to 10 is possible. This is open book exam since test set and wave equation are given

", "rawMarkdown": "I am more optimistic. I even think near to 10 is possible. This is open book exam since test set and wave equation are given", "votes": 2, "replies": [ { "id": 3228604, "postDate": "2025-06-20T10:58:56.563Z", "content": "

congrads to @harshitsheoran !!!!
\nhe is the first to cross the 10 line.

\n

\"\"

\n

i think gold line will be near 10.
\ni am looking foward to breaking 5 line for rank1 winner.
\ngo go go!

\n
\n

why i think so:

\n", "rawMarkdown": "congrads to @harshitsheoran !!!!\nhe is the first to cross the 10 line.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Faa69d58411cab5532aad923cbdfdaf42%2FSelection_999(8254).png?generation=1750416997222079&alt=media)\n\ni think gold line will be near 10.\ni am looking foward to breaking 5 line for rank1 winner.\ngo go go!\n\n---\n\nwhy i think so:\n- data science practitioners, one should be able to tell the performance of data-driven and non data driven approach. statistics has limit but ...", "votes": 3, "replies": [ { "id": 3228610, "postDate": "2025-06-20T11:05:45.323Z", "content": "

Holy **! Really get it??? I can't wait to see the top solution!

", "rawMarkdown": "Holy **! Really get it??? I can't wait to see the top solution!", "votes": 1 }, { "id": 3228619, "postDate": "2025-06-20T11:11:10.410Z", "content": "

Magic of 10 :D, its became like a Santa Competition :D

", "rawMarkdown": "Magic of 10 :D, its became like a Santa Competition :D" }, { "id": 3228989, "postDate": "2025-06-20T19:41:35.197Z", "content": "

note with foward wave code, train error will be zero. that is why i am not surprise public test (and private) can fall below 5. this competition is really about :
\n\"This competition challenges you to bridge the gap by combining physics and machine learning to advance\"

\n
\n

\"Traditional physics-based approaches are accurate, but incredibly slow and prone to errors when the signal is weak from noisy data. Pure machine learning solutions are faster, but require vast amounts of labeled data and often fail to generalize to new, unfamiliar \"signal.\"\"

\n

will even be better if you solution is robust for data may be weak/unfamiliar …

", "rawMarkdown": "note with foward wave code, train error will be zero. that is why i am not surprise public test (and private) can fall below 5. this competition is really about :\n\"This competition challenges you to **bridge the gap by combining physics and machine learning** to advance\"\n\n---\n\n\"Traditional physics-based approaches are accurate, but incredibly slow and prone to errors when the signal is weak from noisy data. Pure machine learning solutions are faster, but require vast amounts of labeled data and often fail to generalize to new, unfamiliar \"signal.\"\"\n\nwill even be better if you solution is robust for data may be weak/unfamiliar ...\n\n", "votes": 2 }, { "id": 3228995, "postDate": "2025-06-20T20:07:21.400Z", "content": "

I think one approach is to use forward wave code to make more train data. That will improve model CV LB compared to using only the data Kaggle has provided.

", "rawMarkdown": "I think one approach is to use forward wave code to make more train data. That will improve model CV LB compared to using only the data Kaggle has provided." }, { "id": 3229029, "postDate": "2025-06-20T21:42:17.280Z", "content": "
\n

using only the data Kaggle has provided.

\n
\n

This leads nowhere as this data is too small. Hosts said to use open-fwi dataset and there are several variants available on kaggle or huggingface.

", "rawMarkdown": "> using only the data Kaggle has provided.\n\nThis leads nowhere as this data is too small. Hosts said to use open-fwi dataset and there are several variants available on kaggle or huggingface.", "votes": 1 }, { "id": 3229215, "postDate": "2025-06-21T07:01:13.040Z", "content": "

I have to privilege to join a few of winning team kaggle meeting in previous competitions, simplicity of top solution always surprises me. As a ml person, I thought data driven is the king. But there is always another king. That is the beauty of kaggle, it let u see the other side of the world. This competition is intriguing because it is the battle of few kings. I am anxious to see the results. I begin to believe that neural net is can never be universal function. It can only be universal approximate function.

", "rawMarkdown": "I have to privilege to join a few of winning team kaggle meeting in previous competitions, simplicity of top solution always surprises me. As a ml person, I thought data driven is the king. But there is always another king. That is the beauty of kaggle, it let u see the other side of the world. This competition is intriguing because it is the battle of few kings. I am anxious to see the results. I begin to believe that neural net is can never be universal function. It can only be universal approximate function.", "votes": 2 } ] } ] }, { "id": 3226999, "postDate": "2025-06-18T10:54:29.287Z", "content": "

Crazy time…

", "rawMarkdown": "Crazy time...", "votes": 1 }, { "id": 3226915, "postDate": "2025-06-18T08:35:52.013Z", "content": "

I do believe that 17.1 will stay in gold medal zone although near the bottom, so maybe 5-6 teams breaks sub 15, call me optimistic :D

", "rawMarkdown": "I do believe that 17.1 will stay in gold medal zone although near the bottom, so maybe 5-6 teams breaks sub 15, call me optimistic :D", "votes": 1, "replies": [ { "id": 3226928, "postDate": "2025-06-18T09:07:13.953Z", "content": "

You make us nervous, i think 19.5 is gold medal line!

", "rawMarkdown": "You make us nervous, i think 19.5 is gold medal line!", "votes": 2, "replies": [ { "id": 3226958, "postDate": "2025-06-18T09:56:45.037Z", "content": "

A week ago, 7th place had a score of like 23 … I am being optimistic :D that such score boost can be gained again

", "rawMarkdown": "A week ago, 7th place had a score of like 23 ... I am being optimistic :D that such score boost can be gained again" }, { "id": 3227642, "postDate": "2025-06-19T06:40:13.673Z", "content": "

You are right, 19.5 is not enough😱😱😱

", "rawMarkdown": "You are right, 19.5 is not enough😱😱😱" }, { "id": 3227651, "postDate": "2025-06-19T06:50:26.690Z", "content": "

So many teams suddenly popping under 20.

", "rawMarkdown": "So many teams suddenly popping under 20." }, { "id": 3230438, "postDate": "2025-06-23T03:42:50.393Z", "content": "

You were right, 19.5 was not enough😱😱😱

", "rawMarkdown": "You were right, 19.5 was not enough😱😱😱" }, { "id": 3230455, "postDate": "2025-06-23T04:09:08.573Z", "content": "

Even 17.1 won't be in gold at this rate 😱

", "rawMarkdown": "Even 17.1 won't be in gold at this rate 😱" }, { "id": 3230477, "postDate": "2025-06-23T05:14:00.317Z", "content": "

a lot of team are waiting for their compute to complete. they already know what to do.
\nso msot team will score higher than now at the end of one week.

\n

as i mentioned in a couple of days before, i am looking at \"i think gold line will be near 10.\". Winner will be less

", "rawMarkdown": "a lot of team are waiting for their compute to complete. they already know what to do.\nso msot team will score higher than now at the end of one week.\n\nas i mentioned in a couple of days before, i am looking at \"i think gold line will be near 10.\". Winner will be less", "votes": 3 }, { "id": 3230523, "postDate": "2025-06-23T06:32:49.353Z", "content": "

You make us nervous, i think 15.0 is pretty safe!

", "rawMarkdown": "You make us nervous, i think 15.0 is pretty safe!", "votes": 1 }, { "id": 3230526, "postDate": "2025-06-23T06:37:11.057Z", "content": "

as data scientist, we don't believe in here-say or there-say.

\n

we need to see evidence and trend.
\nthe best way is to collect the scores of teams for the past history and draw extrapolation curves :)

\n

or look at other similar competitions before.

", "rawMarkdown": "as data scientist, we don't believe in here-say or there-say.\n\nwe need to see evidence and trend.\nthe best way is to collect the scores of teams for the past history and draw extrapolation curves :)\n\nor look at other similar competitions before.\n" }, { "id": 3230535, "postDate": "2025-06-23T06:59:34.923Z", "content": "

I might have seen that curve before somewhere
\nhttps://www.kaggle.com/competitions/waveform-inversion/discussion/579841#3206525

", "rawMarkdown": "I might have seen that curve before somewhere\nhttps://www.kaggle.com/competitions/waveform-inversion/discussion/579841#3206525", "votes": 1 } ] } ] }, { "id": 3226852, "postDate": "2025-06-18T07:00:04.810Z", "content": "

4 teams with public LB <15.0 by the submission deadline

", "rawMarkdown": "4 teams with public LB <15.0 by the submission deadline", "votes": 2, "replies": [ { "id": 3230294, "postDate": "2025-06-22T18:11:50.993Z", "content": "

stop the count!\"\"

", "rawMarkdown": "**stop the count!**![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13977249%2Fa672542536630091adfc5cb10bf27c5a%2Fkaggle.jpg?generation=1750615900620014&alt=media)", "replies": [ { "id": 3230297, "postDate": "2025-06-22T18:14:19.807Z", "content": "

Maybe 4 teams below 10 by deadline? 😂

", "rawMarkdown": "Maybe 4 teams below 10 by deadline? 😂", "votes": 3 }, { "id": 3232963, "postDate": "2025-06-26T10:59:59.500Z", "content": "

I think more than 4 is likely.

", "rawMarkdown": "I think more than 4 is likely." } ] } ] }, { "id": 3226845, "postDate": "2025-06-18T06:51:08.950Z", "content": "

Hard to guess. Including your team?

", "rawMarkdown": "Hard to guess. Including your team?", "votes": 2, "replies": [ { "id": 3226867, "postDate": "2025-06-18T07:18:03.240Z", "content": "

Haha, really do hope so.

", "rawMarkdown": "Haha, really do hope so.", "votes": 2 } ] }, { "id": 3226912, "postDate": "2025-06-18T08:33:11.120Z", "content": "

1 at most @cnumber

", "rawMarkdown": "1 at most @cnumber ", "replies": [ { "id": 3229030, "postDate": "2025-06-20T21:43:14.373Z", "content": "

This did not age well.

", "rawMarkdown": "This did not age well." } ] }, { "id": 3226900, "postDate": "2025-06-18T08:20:01.003Z", "content": "

4-5 Teams - your team will one of it - as your team cracked many 2025 competitions.

", "rawMarkdown": "4-5 Teams - your team will one of it - as your team cracked many 2025 competitions." }, { "id": 3228958, "postDate": "2025-06-20T18:25:35.653Z", "rawMarkdown": "", "isDeleted": true }, { "id": 3227769, "postDate": "2025-06-19T09:01:09.573Z", "rawMarkdown": "", "votes": -2, "isDeleted": true }, { "id": 3227461, "postDate": "2025-06-19T03:01:49.870Z", "rawMarkdown": "", "votes": -5, "isDeleted": true } ], "comments": [ { "id": 3232029, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-25T09:44:46.667000", "content": "

the speed of progress of the leaderboard in the last few days is still shocking to me, even though i expect as dramatic improvement. Based on my understanding of the \"solution\", my conclusion is that many top teams have many very good GPUs. I suppose they are in the range of A100,H100 (maybe some have H200, any have B200?),4090 and 5090. Hard to imagine top team proably have at least [masked] cards ? …

\n

the scale of computation resource today is different from those when i first joined kaggle more than 10 years ago .

\n

\"\"
\nGPU working hard at last days of competition … go go go!

", "votes": 4, "replies": [ { "id": 3232074, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-25T10:41:59.680000", "content": "

Someone said he had spare GB200.

", "votes": 0, "replies": [ { "id": 3233096, "author_name": "", "author_url": "", "post_date": "2025-06-26T13:28:29.383000", "content": "

lol that's me
\nif only I had these a few months ago in AIMO-2 :/

", "votes": 1, "replies": [] } ] }, { "id": 3232077, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-25T10:44:49.413000", "content": "

You say \"solution\" as if it's not a real solution.
\nEven if you know what to do, actually doing it is still not trivial! 🤣

", "votes": 1, "replies": [ { "id": 3232081, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-25T10:49:23.367000", "content": "

of course! the details are in the devil!

\n

everyone knows what to do,
\nbut only one would survives

\n
\n

last time when i was in company, my project manager boss would suggest \"if detection is not good enough, then do tracking to improve the accuracy\"

\n

then in my mind, i would grumble \"you have to tell me how to implement the tracking!!! tracking is not the solution. the details are\"

\n
\n

but then when you explain your solution to your client, you don't want to confuse them with the details and also not to spill out the secret sauce, you would say \" the winning part of our solution when compared to competitors in the marked is that we added tracking to normal detection\".

", "votes": 1, "replies": [] }, { "id": 3232186, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-25T13:23:41.500000", "content": "
\n

everyone knows what to do,

\n
\n

I doubt it. I don't think I do to start with.

", "votes": 0, "replies": [] }, { "id": 3232402, "author_name": "guo dashuai", "author_url": "", "post_date": "2025-06-25T16:50:03.953000", "content": "

Most teams are still calculating—I believe it's just a matter of time before the rest are done.

", "votes": 0, "replies": [] } ] } ] }, { "id": 3227217, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-18T17:01:12.513000", "content": "

I am more optimistic. I even think near to 10 is possible. This is open book exam since test set and wave equation are given

", "votes": 2, "replies": [ { "id": 3228604, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-20T10:58:56.563000", "content": "

congrads to @harshitsheoran !!!!
\nhe is the first to cross the 10 line.

\n

\"\"

\n

i think gold line will be near 10.
\ni am looking foward to breaking 5 line for rank1 winner.
\ngo go go!

\n
\n

why i think so:

\n", "votes": 3, "replies": [ { "id": 3228610, "author_name": "Kurise", "author_url": "", "post_date": "2025-06-20T11:05:45.323000", "content": "

Holy **! Really get it??? I can't wait to see the top solution!

", "votes": 1, "replies": [] }, { "id": 3228619, "author_name": "SeshuRaju 🧘‍♂️", "author_url": "", "post_date": "2025-06-20T11:11:10.410000", "content": "

Magic of 10 :D, its became like a Santa Competition :D

", "votes": 0, "replies": [] }, { "id": 3228989, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-20T19:41:35.197000", "content": "

note with foward wave code, train error will be zero. that is why i am not surprise public test (and private) can fall below 5. this competition is really about :
\n\"This competition challenges you to bridge the gap by combining physics and machine learning to advance\"

\n
\n

\"Traditional physics-based approaches are accurate, but incredibly slow and prone to errors when the signal is weak from noisy data. Pure machine learning solutions are faster, but require vast amounts of labeled data and often fail to generalize to new, unfamiliar \"signal.\"\"

\n

will even be better if you solution is robust for data may be weak/unfamiliar …

", "votes": 2, "replies": [] }, { "id": 3228995, "author_name": "Chris Deotte", "author_url": "", "post_date": "2025-06-20T20:07:21.400000", "content": "

I think one approach is to use forward wave code to make more train data. That will improve model CV LB compared to using only the data Kaggle has provided.

", "votes": 0, "replies": [] }, { "id": 3229029, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-20T21:42:17.280000", "content": "
\n

using only the data Kaggle has provided.

\n
\n

This leads nowhere as this data is too small. Hosts said to use open-fwi dataset and there are several variants available on kaggle or huggingface.

", "votes": 1, "replies": [] }, { "id": 3229215, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-21T07:01:13.040000", "content": "

I have to privilege to join a few of winning team kaggle meeting in previous competitions, simplicity of top solution always surprises me. As a ml person, I thought data driven is the king. But there is always another king. That is the beauty of kaggle, it let u see the other side of the world. This competition is intriguing because it is the battle of few kings. I am anxious to see the results. I begin to believe that neural net is can never be universal function. It can only be universal approximate function.

", "votes": 2, "replies": [] } ] } ] }, { "id": 3226999, "author_name": "Kurise", "author_url": "", "post_date": "2025-06-18T10:54:29.287000", "content": "

Crazy time…

", "votes": 1, "replies": [] }, { "id": 3226915, "author_name": "Harshit Sheoran", "author_url": "", "post_date": "2025-06-18T08:35:52.013000", "content": "

I do believe that 17.1 will stay in gold medal zone although near the bottom, so maybe 5-6 teams breaks sub 15, call me optimistic :D

", "votes": 1, "replies": [ { "id": 3226928, "author_name": "hyd", "author_url": "", "post_date": "2025-06-18T09:07:13.953000", "content": "

You make us nervous, i think 19.5 is gold medal line!

", "votes": 2, "replies": [ { "id": 3226958, "author_name": "Harshit Sheoran", "author_url": "", "post_date": "2025-06-18T09:56:45.037000", "content": "

A week ago, 7th place had a score of like 23 … I am being optimistic :D that such score boost can be gained again

", "votes": 0, "replies": [] }, { "id": 3227642, "author_name": "hyd", "author_url": "", "post_date": "2025-06-19T06:40:13.673000", "content": "

You are right, 19.5 is not enough😱😱😱

", "votes": 0, "replies": [] }, { "id": 3227651, "author_name": "c-number", "author_url": "", "post_date": "2025-06-19T06:50:26.690000", "content": "

So many teams suddenly popping under 20.

", "votes": 0, "replies": [] }, { "id": 3230438, "author_name": "c-number", "author_url": "", "post_date": "2025-06-23T03:42:50.393000", "content": "

You were right, 19.5 was not enough😱😱😱

", "votes": 0, "replies": [] }, { "id": 3230455, "author_name": "Harshit Sheoran", "author_url": "", "post_date": "2025-06-23T04:09:08.573000", "content": "

Even 17.1 won't be in gold at this rate 😱

", "votes": 0, "replies": [] }, { "id": 3230477, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-23T05:14:00.317000", "content": "

a lot of team are waiting for their compute to complete. they already know what to do.
\nso msot team will score higher than now at the end of one week.

\n

as i mentioned in a couple of days before, i am looking at \"i think gold line will be near 10.\". Winner will be less

", "votes": 3, "replies": [] }, { "id": 3230523, "author_name": "hyd", "author_url": "", "post_date": "2025-06-23T06:32:49.353000", "content": "

You make us nervous, i think 15.0 is pretty safe!

", "votes": 1, "replies": [] }, { "id": 3230526, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-23T06:37:11.057000", "content": "

as data scientist, we don't believe in here-say or there-say.

\n

we need to see evidence and trend.
\nthe best way is to collect the scores of teams for the past history and draw extrapolation curves :)

\n

or look at other similar competitions before.

", "votes": 0, "replies": [] }, { "id": 3230535, "author_name": "c-number", "author_url": "", "post_date": "2025-06-23T06:59:34.923000", "content": "

I might have seen that curve before somewhere
\nhttps://www.kaggle.com/competitions/waveform-inversion/discussion/579841#3206525

", "votes": 1, "replies": [] } ] } ] }, { "id": 3226852, "author_name": "Yurnero", "author_url": "", "post_date": "2025-06-18T07:00:04.810000", "content": "

4 teams with public LB <15.0 by the submission deadline

", "votes": 2, "replies": [ { "id": 3230294, "author_name": "Yurnero", "author_url": "", "post_date": "2025-06-22T18:11:50.993000", "content": "

stop the count!\"\"

", "votes": 0, "replies": [ { "id": 3230297, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-22T18:14:19.807000", "content": "

Maybe 4 teams below 10 by deadline? 😂

", "votes": 3, "replies": [] }, { "id": 3232963, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-26T10:59:59.500000", "content": "

I think more than 4 is likely.

", "votes": 0, "replies": [] } ] } ] }, { "id": 3226845, "author_name": "hyd", "author_url": "", "post_date": "2025-06-18T06:51:08.950000", "content": "

Hard to guess. Including your team?

", "votes": 2, "replies": [ { "id": 3226867, "author_name": "c-number", "author_url": "", "post_date": "2025-06-18T07:18:03.240000", "content": "

Haha, really do hope so.

", "votes": 2, "replies": [] } ] }, { "id": 3226912, "author_name": "Ravi Ramakrishnan", "author_url": "", "post_date": "2025-06-18T08:33:11.120000", "content": "

1 at most @cnumber

", "votes": 0, "replies": [ { "id": 3229030, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-20T21:43:14.373000", "content": "

This did not age well.

", "votes": 0, "replies": [] } ] }, { "id": 3226900, "author_name": "SeshuRaju 🧘‍♂️", "author_url": "", "post_date": "2025-06-18T08:20:01.003000", "content": "

4-5 Teams - your team will one of it - as your team cracked many 2025 competitions.

", "votes": 0, "replies": [] }, { "id": 3228958, "author_name": "", "author_url": "", "post_date": "2025-06-20T18:25:35.653000", "content": "", "votes": 0, "replies": [] }, { "id": 3227769, "author_name": "", "author_url": "", "post_date": "2025-06-19T09:01:09.573000", "content": "", "votes": -2, "replies": [] }, { "id": 3227461, "author_name": "", "author_url": "", "post_date": "2025-06-19T03:01:49.870000", "content": "", "votes": -5, "replies": [] } ], "raw_markdown_by_id": { "3226841": "My guess is 3 teams.", "3232029": "the speed of progress of the leaderboard in the last few days is still shocking to me, even though i expect as dramatic improvement. Based on my understanding of the \"solution\", my conclusion is that many top teams have many very good GPUs. I suppose they are in the range of A100,H100 (maybe some have H200, any have B200?),4090 and 5090. Hard to imagine top team proably have at least [masked] cards ? ...\n\nthe scale of computation resource today is different from those when i first joined kaggle more than 10 years ago .\n\n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Feac8dfd9ab51ba3d2424f3540d9e005d%2FSelection_999(8272).png?generation=1750846093211093&alt=media)\nGPU working hard at last days of competition ... go go go!", "3227217": "I am more optimistic. I even think near to 10 is possible. This is open book exam since test set and wave equation are given", "3226999": "Crazy time...", "3226915": "I do believe that 17.1 will stay in gold medal zone although near the bottom, so maybe 5-6 teams breaks sub 15, call me optimistic :D", "3226852": "4 teams with public LB <15.0 by the submission deadline", "3226845": "Hard to guess. Including your team?", "3226912": "1 at most @cnumber ", "3226900": "4-5 Teams - your team will one of it - as your team cracked many 2025 competitions.", "3228958": "", "3227769": "", "3227461": "" } }