{ "id": 586264, "title": "I did it again: submit twice the same sub", "url": "/competitions/waveform-inversion/discussion/586264", "author_name": "CPMP", "post_date": "2025-06-25T18:46:47.096000", "votes": 13, "comment_count": 42, "views": 0, "content": "

I wish there was a hash check on kaggle to detect when we submit twice the same csv file.

\n

I think I created a feature request for it. But nothing happened.

\n

Anyway, don't be like me, be careful, this can happen the last day and could be dramatic.

", "messages": [ { "id": 3232463, "postDate": "2025-06-25T18:46:47.097Z", "content": "

I wish there was a hash check on kaggle to detect when we submit twice the same csv file.

\n

I think I created a feature request for it. But nothing happened.

\n

Anyway, don't be like me, be careful, this can happen the last day and could be dramatic.

", "rawMarkdown": "I wish there was a hash check on kaggle to detect when we submit twice the same csv file.\n\nI think I created a feature request for it. But nothing happened.\n\nAnyway, don't be like me, be careful, this can happen the last day and could be dramatic. ", "votes": 13 }, { "id": 3236973, "postDate": "2025-06-30T19:32:10.627Z", "content": "

Feels good that I am not alone in submitting twice @cpmpml 😆

", "rawMarkdown": "Feels good that I am not alone in submitting twice @cpmpml 😆", "votes": 1, "replies": [ { "id": 3237049, "postDate": "2025-06-30T20:57:22.740Z", "content": "

Good! I feel less lonely too :D

", "rawMarkdown": "Good! I feel less lonely too :D" } ] }, { "id": 3234264, "postDate": "2025-06-27T17:15:03.153Z", "content": "

same problem…😂

", "rawMarkdown": "same problem...😂", "votes": 1 }, { "id": 3232494, "postDate": "2025-06-25T19:33:10.087Z", "content": "

Hahaha the same thing happened to me today! I ran a new version and I thought it finished so I submitted without looking, and then I saw the new version did not finished yet and what I submitted was the old one, again.

", "rawMarkdown": "Hahaha the same thing happened to me today! I ran a new version and I thought it finished so I submitted without looking, and then I saw the new version did not finished yet and what I submitted was the old one, again.", "votes": 2, "replies": [ { "id": 3232541, "postDate": "2025-06-25T20:53:12.637Z", "content": "

I'm in good company then!

", "rawMarkdown": "I'm in good company then!", "replies": [ { "id": 3232552, "postDate": "2025-06-25T21:23:14.823Z", "content": "

The honor is mine! 🤣

", "rawMarkdown": "The honor is mine! 🤣", "votes": 1 }, { "id": 3233041, "postDate": "2025-06-26T12:40:25.970Z", "content": "

@shlomoron
\nI've been meaning to ask you smth. I've been following you since last year's ARIEL competition btw. How do you manage to get medals so often !? 6 medals in 8 competitions is really chad of you. Being a rookie and still in my sophomore year I can use some tips from you. I've done some handy amount of work on NLP and multi-modal areas but yeah I'd like hear anything from you.

", "rawMarkdown": "@shlomoron \nI've been meaning to ask you smth. I've been following you since last year's ARIEL competition btw. How do you manage to get medals so often !? 6 medals in 8 competitions is really chad of you. Being a rookie and still in my sophomore year I can use some tips from you. I've done some handy amount of work on NLP and multi-modal areas but yeah I'd like hear anything from you. " }, { "id": 3233057, "postDate": "2025-06-26T12:53:10.980Z", "content": "

Thx @aneeshmukkamala. Mainly time and dedication, and I really enjoying it. There are much better people tho…ask @cnumber he is the real chad 🤣

\n

Oh also, there were much more that 8 competitions. I just hide the unimportant ones from my profile lol.

", "rawMarkdown": "Thx @aneeshmukkamala. Mainly time and dedication, and I really enjoying it. There are much better people tho...ask @cnumber he is the real chad 🤣 \n\nOh also, there were much more that 8 competitions. I just hide the unimportant ones from my profile lol." }, { "id": 3233093, "postDate": "2025-06-26T13:26:50.633Z", "content": "

If I may ask, are you guys working or doing anything else? Like how do you find time for kaggle? I know it's pretty rewarding if you finish on the podium but I've crashes and burned in all the 4 competitions that I took place so far and tbh looking at the LB is motivating, frustrating, gasping all at the same time. I brainstorm and come up with smth everytime and it doesnt even match the bronze region. Nevertheless I learnt quite a lot and connected with a bunch of new people. @cnumber can I ask.. how do you do it?

", "rawMarkdown": "If I may ask, are you guys working or doing anything else? Like how do you find time for kaggle? I know it's pretty rewarding if you finish on the podium but I've crashes and burned in all the 4 competitions that I took place so far and tbh looking at the LB is motivating, frustrating, gasping all at the same time. I brainstorm and come up with smth everytime and it doesnt even match the bronze region. Nevertheless I learnt quite a lot and connected with a bunch of new people. @cnumber can I ask.. how do you do it? " }, { "id": 3233111, "postDate": "2025-06-26T13:39:16.963Z", "content": "

For every competition, I begin by identifying the optimal solution without writing any code or with a little bit of EDA.
\nFor example in last year's NeurIPS Ariel Data Challenge, it was quite obvious that the winner's team would take the physics behind the data into account by reading the data generation code, which is what we decided to do first.
\nOr for the recent chess competition, a little bit of googling and ChatGPTing would tell you that you had to use stockfish and NNUE to secure gold medal.
\nAfter I'm confident enough with the solution, I write down all the tasks that I believe would lead me to 1st place, and just try them all.

", "rawMarkdown": "For every competition, I begin by identifying the optimal solution without writing any code or with a little bit of EDA.\nFor example in last year's NeurIPS Ariel Data Challenge, it was quite obvious that the winner's team would take the physics behind the data into account by reading the data generation code, which is what we decided to do first.\nOr for the recent chess competition, a little bit of googling and ChatGPTing would tell you that you had to use stockfish and NNUE to secure gold medal.\nAfter I'm confident enough with the solution, I write down all the tasks that I believe would lead me to 1st place, and just try them all.", "votes": 5 }, { "id": 3233121, "postDate": "2025-06-26T13:47:12.643Z", "content": "
\n

it was quite obvious that the winner's team would take the physics behind the data into account by reading the data generation code, which is what we decided to do first.

\n
\n

It was actually very very far from obvious 🤣 how many teams did it? This is why you are genius.

", "rawMarkdown": "> it was quite obvious that the winner's team would take the physics behind the data into account by reading the data generation code, which is what we decided to do first. \n\nIt was actually very very far from obvious 🤣 how many teams did it? This is why you are genius.", "votes": 1 }, { "id": 3233181, "postDate": "2025-06-26T14:40:19.987Z", "rawMarkdown": "", "isDeleted": true }, { "id": 3233702, "postDate": "2025-06-27T04:50:51.017Z", "content": "

By the way, will you be joining this year's Ariel competition? @shlomoron
\nI have already teamed up with daiwakun to secure 1st prize :)

", "rawMarkdown": "By the way, will you be joining this year's Ariel competition? @shlomoron\nI have already teamed up with daiwakun to secure 1st prize :)" }, { "id": 3233709, "postDate": "2025-06-27T04:57:34.087Z", "content": "

actually you can start a tutorial on 2024 competition.
\ni think many kagglers will be interested.
\nsince it is past competitions tutorial, there is no private sharing if you don't discuss 2025 at all.

", "rawMarkdown": "actually you can start a tutorial on 2024 competition.\ni think many kagglers will be interested.\nsince it is past competitions tutorial, there is no private sharing if you don't discuss 2025 at all.\n\n" }, { "id": 3233728, "postDate": "2025-06-27T05:20:18.153Z", "content": "

Good idea!
\nThe presentations from last year's NeurIPS Workshop are avaible online, so I decided to share them.
\nhttps://www.kaggle.com/competitions/ariel-data-challenge-2025/discussion/586548

", "rawMarkdown": "Good idea!\nThe presentations from last year's NeurIPS Workshop are avaible online, so I decided to share them.\nhttps://www.kaggle.com/competitions/ariel-data-challenge-2025/discussion/586548", "votes": 1 }, { "id": 3233798, "postDate": "2025-06-27T07:18:13.033Z", "content": "
\n

By the way, will you be joining this year's Ariel competition? @shlomoron
\n I have already teamed up with daiwakun to secure 1st prize :)

\n
\n

Probably, after a couple of days of resting from this one 🤣
\nI think I'll try to team up with someone, too, this time around. There are a couple of interesting competitions, so getting a solo price range will be difficult. We'll see.

", "rawMarkdown": ">By the way, will you be joining this year's Ariel competition? @shlomoron\nI have already teamed up with daiwakun to secure 1st prize :) \n\nProbably, after a couple of days of resting from this one 🤣 \nI think I'll try to team up with someone, too, this time around. There are a couple of interesting competitions, so getting a solo price range will be difficult. We'll see." }, { "id": 3233823, "postDate": "2025-06-27T07:55:32.153Z", "content": "

\"resting from this one \"
\nCorrected:
\n\"after the GPUs resting from this one \"

", "rawMarkdown": "\"resting from this one \"\nCorrected:\n\"after the GPUs resting from this one \"\n" }, { "id": 3233828, "postDate": "2025-06-27T08:07:47.297Z", "content": "
\n

\"after the GPUs resting from this one \"

\n
\n

No need, I use cloud. They can all burn afaic, it's not my problem :)

", "rawMarkdown": ">\"after the GPUs resting from this one \" \n\nNo need, I use cloud. They can all burn afaic, it's not my problem :)" }, { "id": 3233833, "postDate": "2025-06-27T08:14:51.600Z", "content": "

@shlomoron
\nCan we team up 🙃!?

", "rawMarkdown": "@shlomoron \nCan we team up 🙃!? " }, { "id": 3233845, "postDate": "2025-06-27T08:26:52.637Z", "content": "
\n

@shlomoron
\n Can we team up 🙃!?

\n
\n

For Ariel? Be solo price range on the public LB a month and a half into it, then maybe, if I wouldn't be in a team yet 😅

", "rawMarkdown": ">@shlomoron\nCan we team up 🙃!? \n\nFor Ariel? Be solo price range on the public LB a month and a half into it, then maybe, if I wouldn't be in a team yet 😅", "votes": 1 }, { "id": 3233853, "postDate": "2025-06-27T08:48:21.283Z", "rawMarkdown": "", "isDeleted": true }, { "id": 3234073, "postDate": "2025-06-27T13:36:11.180Z", "content": "
\n

By the way, will you be joining this year's Ariel competition? @shlomoron
\n I have already teamed up with daiwakun to secure 1st prize :)

\n
\n

We'll see about that.

", "rawMarkdown": "> By the way, will you be joining this year's Ariel competition? @shlomoron\n> I have already teamed up with daiwakun to secure 1st prize :)\n\nWe'll see about that.", "votes": 1 }, { "id": 3234084, "postDate": "2025-06-27T13:44:32.380Z", "content": "
\n

We'll see about that.

\n
\n

If you three join together it will really be secure! 🤣

", "rawMarkdown": ">We'll see about that. \n\nIf you three join together it will really be secure! 🤣" }, { "id": 3234107, "postDate": "2025-06-27T14:14:01Z", "content": "

If you two team up, our team is in huge trouble.

", "rawMarkdown": "If you two team up, our team is in huge trouble." }, { "id": 3234144, "postDate": "2025-06-27T14:50:24.710Z", "content": "

can anyone help me out with something. I noticed something in the Qwen3 models.
\nwith no one responding in the AIMO discussions I thought I could use someone's view on this who have dealt with these things before.

\n
 transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \n\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nprompt = \nmessages = [\n    {: , : prompt},\n    # {:  , : }\n]\n\nmodes = [ , ]\n mode  modes:\n\n    text = tokenizer.apply_chat_template(\n        messages,\n        =,\n        =,\n        =mode \n    )\n\n    ()\n    (mode)\n    (text)\n
\n

Should'nt the think tokens be in the enable_thinking thing?

\n
\n\n<|im_start|>\nGive me a short introduction    model.<|im_end|>\n<|im_start|>assistant\n\n\n\n<|im_start|>\nGive me a short introduction    model.<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n
\n

@shlomoron any help would be much appreciated.

", "rawMarkdown": "can anyone help me out with something. I noticed something in the Qwen3 models.\nwith no one responding in the AIMO discussions I thought I could use someone's view on this who have dealt with these things before. \n\n```\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \"Qwen/Qwen3-8B\"\n\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nprompt = \"Give me a short introduction to large language model.\"\nmessages = [\n {\"role\": \"user\", \"content\": prompt},\n # {\"role\": \"system\" , \"content\": \"hi\"}\n]\n\nmodes = [True , False]\nfor mode in modes:\n \n text = tokenizer.apply_chat_template(\n messages,\n tokenize=False,\n add_generation_prompt=True,\n enable_thinking=mode \n )\n\n print(\"-\"*100)\n print(mode)\n print(text)\n```\n\nShould'nt the think tokens be in the enable_thinking thing?\n\n\n```\n----------------------------------------------------------------------------------------------------\nTrue\n<|im_start|>user\nGive me a short introduction to large language model.<|im_end|>\n<|im_start|>assistant\n\n----------------------------------------------------------------------------------------------------\nFalse\n<|im_start|>user\nGive me a short introduction to large language model.<|im_end|>\n<|im_start|>assistant\n\n\n\n\n```\n\n@shlomoron any help would be much appreciated. \n" }, { "id": 3234151, "postDate": "2025-06-27T14:57:37.257Z", "content": "

I don't understand the question. The output looks correct.
\nSee here.

", "rawMarkdown": "I don't understand the question. The output looks correct. \nSee [here](https://huggingface.co/Qwen/Qwen3-8B#switching-between-thinking-and-non-thinking-mode)." }, { "id": 3234170, "postDate": "2025-06-27T15:19:51.633Z", "content": "

I mean the enable thinking means some reasoning is being done/trigerred right? The < think > tokens are'nt present when I flip the enable_thinking to True, should'n these tokens be present? I'm saying this based on the tokenized text I printed.

", "rawMarkdown": "I mean the enable thinking means some reasoning is being done/trigerred right? The < think > tokens are'nt present when I flip the enable_thinking to True, should'n these tokens be present? I'm saying this based on the tokenized text I printed. " }, { "id": 3234176, "postDate": "2025-06-27T15:27:50.823Z", "content": "

It's explained in the link.

", "rawMarkdown": "It's explained in the link." }, { "id": 3234183, "postDate": "2025-06-27T15:45:55.503Z", "content": "

Yeah no I got what you're saying. I read that too.
\nThis bit especially

\n
In  mode, the model will   any think content  will   a <think>...</think> block.\n
\n

Like my doubt was that if the < think > blocks are'nt present in its content (no reasoning). Then like why does the input tokenzied text has the < think > tokens when it is set to enable_thinking=False --> (no reasoning, yet there is presence of the < think > tokens in the tokenized text). I mean if the < think > tokens are absent in the \"tokenzied text after applying the tokenizer.apply_chat_template when enable_thinking=True then what's causing the reasoning to be triggered?

\n

Maybe I am not putting my doubt in the most explainable way, but thanks for taking out some time for looking it up. Appreciate it mate.

", "rawMarkdown": "Yeah no I got what you're saying. I read that too. \nThis bit especially\n\n```\nIn this mode, the model will not generate any think content and will not include a ... block.\n```\n\nLike my doubt was that if the < think > blocks are'nt present in its content (no reasoning). Then like why does the input tokenzied text has the < think > tokens when it is set to ``` enable_thinking=False``` --> (no reasoning, yet there is presence of the < think > tokens in the tokenized text). I mean if the < think > tokens are absent in the \"tokenzied text after applying the ```tokenizer.apply_chat_template``` when ```enable_thinking=True``` then what's causing the reasoning to be triggered? \n\nMaybe I am not putting my doubt in the most explainable way, but thanks for taking out some time for looking it up. Appreciate it mate. " }, { "id": 3234198, "postDate": "2025-06-27T15:53:49.453Z", "content": "

When it is false, the template inserts an empty thinking block.
\nWhen the model sees a thinking block, it generates the final answer without generating more thinking, even if it is an empty block.
\nThis is why empty thinking blocks = no thinking mode.

", "rawMarkdown": "When it is false, the template inserts an empty thinking block. \nWhen the model sees a thinking block, it generates the final answer without generating more thinking, even if it is an empty block. \nThis is why empty thinking blocks = no thinking mode.\n" }, { "id": 3234219, "postDate": "2025-06-27T16:16:33.557Z", "rawMarkdown": "", "isDeleted": true }, { "id": 3234480, "postDate": "2025-06-27T22:27:25.620Z", "content": "

there are so many interesting competitions. i am interested in both NeurIPS 2025- Open Polymer and Ariel competition. maybe we need cross competition team collaborations, different team members focusing on different competitions and report on progress to the whole team.. … just like doing multiple projects in company.

", "rawMarkdown": "there are so many interesting competitions. i am interested in both NeurIPS 2025- Open Polymer and Ariel competition. maybe we need cross competition team collaborations, different team members focusing on different competitions and report on progress to the whole team.. ... just like doing multiple projects in company." }, { "id": 3234485, "postDate": "2025-06-27T22:44:28.547Z", "content": "
\n

i am interested in both NeurIPS 2025- Open Polymer and Ariel competition

\n
\n

Open polymer looks too similar to BELKA. Do you think it will be any better lottery-wise?

", "rawMarkdown": ">i am interested in both NeurIPS 2025- Open Polymer and Ariel competition \n\n\nOpen polymer looks too similar to BELKA. Do you think it will be any better lottery-wise? " }, { "id": 3234530, "postDate": "2025-06-28T02:31:35.127Z", "content": "

i haven't check yest (too busy with current wave inversion … )

\n

but i do have some ideas to try for molecuar property prediction after reading recent works from IBM,microsoft,etc … they claim to have very good foundation models, etc to take care of OOD.

\n

hopefull, from cross competition team from past winners, one can learn from each other the drak magic and secrets.

", "rawMarkdown": "i haven't check yest (too busy with current wave inversion ... )\n\nbut i do have some ideas to try for molecuar property prediction after reading recent works from IBM,microsoft,etc ... they claim to have very good foundation models, etc to take care of OOD.\n\nhopefull, from cross competition team from past winners, one can learn from each other the drak magic and secrets." } ] } ] }, { "id": 3233664, "postDate": "2025-06-27T03:50:46.120Z", "content": "

i also meet this problem.

", "rawMarkdown": "i also meet this problem.", "votes": 1 }, { "id": 3232492, "postDate": "2025-06-25T19:21:21.020Z", "content": "

Name them differently.

", "rawMarkdown": "Name them differently.", "replies": [ { "id": 3232542, "postDate": "2025-06-25T20:54:19.523Z", "content": "

They are named differently. I have no excuse.

", "rawMarkdown": "They are named differently. I have no excuse.", "votes": 2 } ] }, { "id": 3232483, "postDate": "2025-06-25T19:09:13.873Z", "content": "

i submit via kaggle command \"kaggle competitions submit\" …using python os system. i think there is a command (?) to list all submissions made, then you can use python to check.

\n

if there is no such command, you make a script to record submission to a history or log locally.
\nmaybe can ask chatgpt to make a gui for that and submit for future LLM app kaggle copmetition

", "rawMarkdown": "i submit via ~~jaggle~~ kaggle command \"kaggle competitions submit\" ...using python os system. i think there is a command (?) to list all submissions made, then you can use python to check.\n\nif there is no such command, you make a script to record submission to a history or log locally.\nmaybe can ask chatgpt to make a gui for that and submit for future LLM app kaggle copmetition", "replies": [ { "id": 3232543, "postDate": "2025-06-25T20:55:31.913Z", "content": "

what is jaggle?

\n

I use kaggle cli to submit.

", "rawMarkdown": "what is jaggle?\n\nI use kaggle cli to submit.", "replies": [ { "id": 3232780, "postDate": "2025-06-26T07:43:16.560Z", "content": "

For those who downvote, there is an ai platform called jaggle that claims to support Ai projects. I searched for a kaggle command in it and could not find it. https://jaggle.ai/

", "rawMarkdown": "For those who downvote, there is an ai platform called jaggle that claims to support Ai projects. I searched for a kaggle command in it and could not find it. https://jaggle.ai/" } ] }, { "id": 3232796, "postDate": "2025-06-26T08:19:28.593Z", "content": "

sorry for the confusion. i first use 'jaggle' in my post. it was a typo error and i corrected it.

", "rawMarkdown": "sorry for the confusion. i first use 'jaggle' in my post. it was a typo error and i corrected it.", "votes": 1, "replies": [ { "id": 3232815, "postDate": "2025-06-26T08:40:08.330Z", "content": "

Sorry, I should have thought of it. But I saw jaggle.ai which made me think you used it.

\n

TL;DR I use kaggle cli and it did not prevent me from making my blunder.

", "rawMarkdown": "Sorry, I should have thought of it. But I saw jaggle.ai which made me think you used it.\n\nTL;DR I use kaggle cli and it did not prevent me from making my blunder.", "votes": 1 } ] } ] } ], "comments": [ { "id": 3236973, "author_name": "Pai", "author_url": "", "post_date": "2025-06-30T19:32:10.627000", "content": "

Feels good that I am not alone in submitting twice @cpmpml 😆

", "votes": 1, "replies": [ { "id": 3237049, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-30T20:57:22.740000", "content": "

Good! I feel less lonely too :D

", "votes": 0, "replies": [] } ] }, { "id": 3234264, "author_name": "", "author_url": "", "post_date": "2025-06-27T17:15:03.153000", "content": "

same problem…😂

", "votes": 1, "replies": [] }, { "id": 3232494, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-25T19:33:10.087000", "content": "

Hahaha the same thing happened to me today! I ran a new version and I thought it finished so I submitted without looking, and then I saw the new version did not finished yet and what I submitted was the old one, again.

", "votes": 2, "replies": [ { "id": 3232541, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-25T20:53:12.637000", "content": "

I'm in good company then!

", "votes": 0, "replies": [ { "id": 3232552, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-25T21:23:14.823000", "content": "

The honor is mine! 🤣

", "votes": 1, "replies": [] }, { "id": 3233041, "author_name": "", "author_url": "", "post_date": "2025-06-26T12:40:25.970000", "content": "

@shlomoron
\nI've been meaning to ask you smth. I've been following you since last year's ARIEL competition btw. How do you manage to get medals so often !? 6 medals in 8 competitions is really chad of you. Being a rookie and still in my sophomore year I can use some tips from you. I've done some handy amount of work on NLP and multi-modal areas but yeah I'd like hear anything from you.

", "votes": 0, "replies": [] }, { "id": 3233057, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-26T12:53:10.980000", "content": "

Thx @aneeshmukkamala. Mainly time and dedication, and I really enjoying it. There are much better people tho…ask @cnumber he is the real chad 🤣

\n

Oh also, there were much more that 8 competitions. I just hide the unimportant ones from my profile lol.

", "votes": 0, "replies": [] }, { "id": 3233093, "author_name": "", "author_url": "", "post_date": "2025-06-26T13:26:50.633000", "content": "

If I may ask, are you guys working or doing anything else? Like how do you find time for kaggle? I know it's pretty rewarding if you finish on the podium but I've crashes and burned in all the 4 competitions that I took place so far and tbh looking at the LB is motivating, frustrating, gasping all at the same time. I brainstorm and come up with smth everytime and it doesnt even match the bronze region. Nevertheless I learnt quite a lot and connected with a bunch of new people. @cnumber can I ask.. how do you do it?

", "votes": 0, "replies": [] }, { "id": 3233111, "author_name": "c-number", "author_url": "", "post_date": "2025-06-26T13:39:16.963000", "content": "

For every competition, I begin by identifying the optimal solution without writing any code or with a little bit of EDA.
\nFor example in last year's NeurIPS Ariel Data Challenge, it was quite obvious that the winner's team would take the physics behind the data into account by reading the data generation code, which is what we decided to do first.
\nOr for the recent chess competition, a little bit of googling and ChatGPTing would tell you that you had to use stockfish and NNUE to secure gold medal.
\nAfter I'm confident enough with the solution, I write down all the tasks that I believe would lead me to 1st place, and just try them all.

", "votes": 5, "replies": [] }, { "id": 3233121, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-26T13:47:12.643000", "content": "
\n

it was quite obvious that the winner's team would take the physics behind the data into account by reading the data generation code, which is what we decided to do first.

\n
\n

It was actually very very far from obvious 🤣 how many teams did it? This is why you are genius.

", "votes": 1, "replies": [] }, { "id": 3233181, "author_name": "", "author_url": "", "post_date": "2025-06-26T14:40:19.987000", "content": "", "votes": 0, "replies": [] }, { "id": 3233702, "author_name": "c-number", "author_url": "", "post_date": "2025-06-27T04:50:51.017000", "content": "

By the way, will you be joining this year's Ariel competition? @shlomoron
\nI have already teamed up with daiwakun to secure 1st prize :)

", "votes": 0, "replies": [] }, { "id": 3233709, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-27T04:57:34.087000", "content": "

actually you can start a tutorial on 2024 competition.
\ni think many kagglers will be interested.
\nsince it is past competitions tutorial, there is no private sharing if you don't discuss 2025 at all.

", "votes": 0, "replies": [] }, { "id": 3233728, "author_name": "c-number", "author_url": "", "post_date": "2025-06-27T05:20:18.153000", "content": "

Good idea!
\nThe presentations from last year's NeurIPS Workshop are avaible online, so I decided to share them.
\nhttps://www.kaggle.com/competitions/ariel-data-challenge-2025/discussion/586548

", "votes": 1, "replies": [] }, { "id": 3233798, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T07:18:13.033000", "content": "
\n

By the way, will you be joining this year's Ariel competition? @shlomoron
\n I have already teamed up with daiwakun to secure 1st prize :)

\n
\n

Probably, after a couple of days of resting from this one 🤣
\nI think I'll try to team up with someone, too, this time around. There are a couple of interesting competitions, so getting a solo price range will be difficult. We'll see.

", "votes": 0, "replies": [] }, { "id": 3233823, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-27T07:55:32.153000", "content": "

\"resting from this one \"
\nCorrected:
\n\"after the GPUs resting from this one \"

", "votes": 0, "replies": [] }, { "id": 3233828, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T08:07:47.297000", "content": "
\n

\"after the GPUs resting from this one \"

\n
\n

No need, I use cloud. They can all burn afaic, it's not my problem :)

", "votes": 0, "replies": [] }, { "id": 3233833, "author_name": "", "author_url": "", "post_date": "2025-06-27T08:14:51.600000", "content": "

@shlomoron
\nCan we team up 🙃!?

", "votes": 0, "replies": [] }, { "id": 3233845, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T08:26:52.637000", "content": "
\n

@shlomoron
\n Can we team up 🙃!?

\n
\n

For Ariel? Be solo price range on the public LB a month and a half into it, then maybe, if I wouldn't be in a team yet 😅

", "votes": 1, "replies": [] }, { "id": 3233853, "author_name": "", "author_url": "", "post_date": "2025-06-27T08:48:21.283000", "content": "", "votes": 0, "replies": [] }, { "id": 3234073, "author_name": "Jeroen Cottaar", "author_url": "", "post_date": "2025-06-27T13:36:11.180000", "content": "
\n

By the way, will you be joining this year's Ariel competition? @shlomoron
\n I have already teamed up with daiwakun to secure 1st prize :)

\n
\n

We'll see about that.

", "votes": 1, "replies": [] }, { "id": 3234084, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T13:44:32.380000", "content": "
\n

We'll see about that.

\n
\n

If you three join together it will really be secure! 🤣

", "votes": 0, "replies": [] }, { "id": 3234107, "author_name": "c-number", "author_url": "", "post_date": "2025-06-27T14:14:01", "content": "

If you two team up, our team is in huge trouble.

", "votes": 0, "replies": [] }, { "id": 3234144, "author_name": "", "author_url": "", "post_date": "2025-06-27T14:50:24.710000", "content": "

can anyone help me out with something. I noticed something in the Qwen3 models.
\nwith no one responding in the AIMO discussions I thought I could use someone's view on this who have dealt with these things before.

\n
 transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \n\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nprompt = \nmessages = [\n    {: , : prompt},\n    # {:  , : }\n]\n\nmodes = [ , ]\n mode  modes:\n\n    text = tokenizer.apply_chat_template(\n        messages,\n        =,\n        =,\n        =mode \n    )\n\n    ()\n    (mode)\n    (text)\n
\n

Should'nt the think tokens be in the enable_thinking thing?

\n
\n\n<|im_start|>\nGive me a short introduction    model.<|im_end|>\n<|im_start|>assistant\n\n\n\n<|im_start|>\nGive me a short introduction    model.<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n
\n

@shlomoron any help would be much appreciated.

", "votes": 0, "replies": [] }, { "id": 3234151, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T14:57:37.257000", "content": "

I don't understand the question. The output looks correct.
\nSee here.

", "votes": 0, "replies": [] }, { "id": 3234170, "author_name": "", "author_url": "", "post_date": "2025-06-27T15:19:51.633000", "content": "

I mean the enable thinking means some reasoning is being done/trigerred right? The < think > tokens are'nt present when I flip the enable_thinking to True, should'n these tokens be present? I'm saying this based on the tokenized text I printed.

", "votes": 0, "replies": [] }, { "id": 3234176, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T15:27:50.823000", "content": "

It's explained in the link.

", "votes": 0, "replies": [] }, { "id": 3234183, "author_name": "", "author_url": "", "post_date": "2025-06-27T15:45:55.503000", "content": "

Yeah no I got what you're saying. I read that too.
\nThis bit especially

\n
In  mode, the model will   any think content  will   a <think>...</think> block.\n
\n

Like my doubt was that if the < think > blocks are'nt present in its content (no reasoning). Then like why does the input tokenzied text has the < think > tokens when it is set to enable_thinking=False --> (no reasoning, yet there is presence of the < think > tokens in the tokenized text). I mean if the < think > tokens are absent in the \"tokenzied text after applying the tokenizer.apply_chat_template when enable_thinking=True then what's causing the reasoning to be triggered?

\n

Maybe I am not putting my doubt in the most explainable way, but thanks for taking out some time for looking it up. Appreciate it mate.

", "votes": 0, "replies": [] }, { "id": 3234198, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T15:53:49.453000", "content": "

When it is false, the template inserts an empty thinking block.
\nWhen the model sees a thinking block, it generates the final answer without generating more thinking, even if it is an empty block.
\nThis is why empty thinking blocks = no thinking mode.

", "votes": 0, "replies": [] }, { "id": 3234219, "author_name": "", "author_url": "", "post_date": "2025-06-27T16:16:33.557000", "content": "", "votes": 0, "replies": [] }, { "id": 3234480, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-27T22:27:25.620000", "content": "

there are so many interesting competitions. i am interested in both NeurIPS 2025- Open Polymer and Ariel competition. maybe we need cross competition team collaborations, different team members focusing on different competitions and report on progress to the whole team.. … just like doing multiple projects in company.

", "votes": 0, "replies": [] }, { "id": 3234485, "author_name": "greySnow", "author_url": "", "post_date": "2025-06-27T22:44:28.547000", "content": "
\n

i am interested in both NeurIPS 2025- Open Polymer and Ariel competition

\n
\n

Open polymer looks too similar to BELKA. Do you think it will be any better lottery-wise?

", "votes": 0, "replies": [] }, { "id": 3234530, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-28T02:31:35.127000", "content": "

i haven't check yest (too busy with current wave inversion … )

\n

but i do have some ideas to try for molecuar property prediction after reading recent works from IBM,microsoft,etc … they claim to have very good foundation models, etc to take care of OOD.

\n

hopefull, from cross competition team from past winners, one can learn from each other the drak magic and secrets.

", "votes": 0, "replies": [] } ] } ] }, { "id": 3233664, "author_name": "Mike_Yu_FIFA", "author_url": "", "post_date": "2025-06-27T03:50:46.120000", "content": "

i also meet this problem.

", "votes": 1, "replies": [] }, { "id": 3232492, "author_name": "Ángel Jacinto Sánchez Ruiz", "author_url": "", "post_date": "2025-06-25T19:21:21.020000", "content": "

Name them differently.

", "votes": 0, "replies": [ { "id": 3232542, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-25T20:54:19.523000", "content": "

They are named differently. I have no excuse.

", "votes": 2, "replies": [] } ] }, { "id": 3232483, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-25T19:09:13.873000", "content": "

i submit via kaggle command \"kaggle competitions submit\" …using python os system. i think there is a command (?) to list all submissions made, then you can use python to check.

\n

if there is no such command, you make a script to record submission to a history or log locally.
\nmaybe can ask chatgpt to make a gui for that and submit for future LLM app kaggle copmetition

", "votes": 0, "replies": [ { "id": 3232543, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-25T20:55:31.913000", "content": "

what is jaggle?

\n

I use kaggle cli to submit.

", "votes": 0, "replies": [ { "id": 3232780, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-26T07:43:16.560000", "content": "

For those who downvote, there is an ai platform called jaggle that claims to support Ai projects. I searched for a kaggle command in it and could not find it. https://jaggle.ai/

", "votes": 0, "replies": [] } ] }, { "id": 3232796, "author_name": "hengck23", "author_url": "", "post_date": "2025-06-26T08:19:28.593000", "content": "

sorry for the confusion. i first use 'jaggle' in my post. it was a typo error and i corrected it.

", "votes": 1, "replies": [ { "id": 3232815, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-26T08:40:08.330000", "content": "

Sorry, I should have thought of it. But I saw jaggle.ai which made me think you used it.

\n

TL;DR I use kaggle cli and it did not prevent me from making my blunder.

", "votes": 1, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3232463": "I wish there was a hash check on kaggle to detect when we submit twice the same csv file.\n\nI think I created a feature request for it. But nothing happened.\n\nAnyway, don't be like me, be careful, this can happen the last day and could be dramatic. ", "3236973": "Feels good that I am not alone in submitting twice @cpmpml 😆", "3234264": "same problem...😂", "3232494": "Hahaha the same thing happened to me today! I ran a new version and I thought it finished so I submitted without looking, and then I saw the new version did not finished yet and what I submitted was the old one, again.", "3233664": "i also meet this problem.", "3232492": "Name them differently.", "3232483": "i submit via ~~jaggle~~ kaggle command \"kaggle competitions submit\" ...using python os system. i think there is a command (?) to list all submissions made, then you can use python to check.\n\nif there is no such command, you make a script to record submission to a history or log locally.\nmaybe can ask chatgpt to make a gui for that and submit for future LLM app kaggle copmetition" } }