The competition is over, but this is where the real work comes in for us. Thank you for all your hard work, i will share a longer post later on our thoughts as organisers, but for now we would love to make a quick call:
\nAs organisers we would love to know your approaches, your little tweaks to improve your scores, your eureka moment that changes your ranking in the leaderboard forever. everything!
\nAnd most importantly, share them! You might think your solution is similar to other participants, it might not, but it might not, and we want to each of them, to see if we can find a pattern.
\nWe will also be circulating an exit survey later, stay tuned.
", "messages": [ { "id": 3033687, "postDate": "2024-11-01T12:39:54.967Z", "content": "The competition is over, but this is where the real work comes in for us. Thank you for all your hard work, i will share a longer post later on our thoughts as organisers, but for now we would love to make a quick call:
\nAs organisers we would love to know your approaches, your little tweaks to improve your scores, your eureka moment that changes your ranking in the leaderboard forever. everything!
\nAnd most importantly, share them! You might think your solution is similar to other participants, it might not, but it might not, and we want to each of them, to see if we can find a pattern.
\nWe will also be circulating an exit survey later, stay tuned.
", "rawMarkdown": "The competition is over, but this is where the real work comes in for us. Thank you for all your hard work, i will share a longer post later on our thoughts as organisers, but for now we would love to make a quick call:\n\nAs organisers we would love to know your approaches, your little tweaks to improve your scores, your eureka moment that changes your ranking in the leaderboard forever. everything!\n\nAnd most importantly, share them! You might think your solution is similar to other participants, it might not, but it might not, and we want to each of them, to see if we can find a pattern. \n\nWe will also be circulating an exit survey later, stay tuned. ", "votes": 14 }, { "id": 3033735, "postDate": "2024-11-01T13:14:03.160Z", "content": "Thank you for being a great host of an exciting competition @gordonyip ! Throughout the whole competition you answered all questions from the participants and were always clear about the competition setup. Especially with the distribution shift from train to test.
\nCompetitions with synthetic datasets are historically difficult w.r.t. setup and keeping them leak free. And this one has been one of the best!
", "rawMarkdown": "Thank you for being a great host of an exciting competition @gordonyip ! Throughout the whole competition you answered all questions from the participants and were always clear about the competition setup. Especially with the distribution shift from train to test.\n\nCompetitions with synthetic datasets are historically difficult w.r.t. setup and keeping them leak free. And this one has been one of the best!", "votes": 11, "replies": [ { "id": 3033768, "postDate": "2024-11-01T13:45:08.477Z", "content": "+1 to that, it's not trivial for a competition to have such a responsive and helpful host like you, @gordonyip, that invests so much time in answering questions.
\nAlthough sadly, the most interesting question for me was not answered, which was what was the SOTA prior to the competition in terms of LB score- maybe now you can answer it? ๐๐
Question was here.
Hi @shlomoron - I realise I havent answered that question. It is a very valid question - one that I am also interested to know. However, I am (we are) not able to answer it, because the official Ariel pipeline has not been developed yet. That is part of a larger work that scientists involving in the mission are doing and it is running on a different time scale than this data challenge. Even though we do know somehow what is going to be inside the official pipeline, we cannot use prototypes to run the challenge at the moment as the performance might be not representative and this would violate non disclosure agreements between the involved institutions.
", "rawMarkdown": "Hi @shlomoron - I realise I havent answered that question. It is a very valid question - one that I am also interested to know. However, I am (we are) not able to answer it, because the official Ariel pipeline has not been developed yet. That is part of a larger work that scientists involving in the mission are doing and it is running on a different time scale than this data challenge. Even though we do know somehow what is going to be inside the official pipeline, we cannot use prototypes to run the challenge at the moment as the performance might be not representative and this would violate non disclosure agreements between the involved institutions.", "votes": 4 } ] } ] }, { "id": 3033766, "postDate": "2024-11-01T13:42:11.507Z", "content": "Thank you for the interesting competition!
\nWe are currently working on summarizing the solution, with sum late subs to examin the effect of some of the components.
\nThe full details should be available within the next few days at the latest!
\n(As some of you may have noticed, we've been moving up the leaderboard quite rapidly over the past two days, and even we arenโt entirely sure yet what made our solution so strong!)
Yes, yours was an insane series of jumps; I think all of us are looking forward to finding out how you managed to clench 1st place in the last second
", "rawMarkdown": "Yes, yours was an insane series of jumps; I think all of us are looking forward to finding out how you managed to clench 1st place in the last second ", "votes": 1 } ] } ], "comments": [ { "id": 3033735, "author_name": "Pascal Pfeiffer", "author_url": "", "post_date": "2024-11-01T13:14:03.160000", "content": "Thank you for being a great host of an exciting competition @gordonyip ! Throughout the whole competition you answered all questions from the participants and were always clear about the competition setup. Especially with the distribution shift from train to test.
\nCompetitions with synthetic datasets are historically difficult w.r.t. setup and keeping them leak free. And this one has been one of the best!
", "votes": 11, "replies": [ { "id": 3033768, "author_name": "greySnow", "author_url": "", "post_date": "2024-11-01T13:45:08.477000", "content": "+1 to that, it's not trivial for a competition to have such a responsive and helpful host like you, @gordonyip, that invests so much time in answering questions.
\nAlthough sadly, the most interesting question for me was not answered, which was what was the SOTA prior to the competition in terms of LB score- maybe now you can answer it? ๐๐
Question was here.
Hi @shlomoron - I realise I havent answered that question. It is a very valid question - one that I am also interested to know. However, I am (we are) not able to answer it, because the official Ariel pipeline has not been developed yet. That is part of a larger work that scientists involving in the mission are doing and it is running on a different time scale than this data challenge. Even though we do know somehow what is going to be inside the official pipeline, we cannot use prototypes to run the challenge at the moment as the performance might be not representative and this would violate non disclosure agreements between the involved institutions.
", "votes": 4, "replies": [] } ] } ] }, { "id": 3033766, "author_name": "c-number", "author_url": "", "post_date": "2024-11-01T13:42:11.507000", "content": "Thank you for the interesting competition!
\nWe are currently working on summarizing the solution, with sum late subs to examin the effect of some of the components.
\nThe full details should be available within the next few days at the latest!
\n(As some of you may have noticed, we've been moving up the leaderboard quite rapidly over the past two days, and even we arenโt entirely sure yet what made our solution so strong!)
Yes, yours was an insane series of jumps; I think all of us are looking forward to finding out how you managed to clench 1st place in the last second
", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "3033687": "The competition is over, but this is where the real work comes in for us. Thank you for all your hard work, i will share a longer post later on our thoughts as organisers, but for now we would love to make a quick call:\n\nAs organisers we would love to know your approaches, your little tweaks to improve your scores, your eureka moment that changes your ranking in the leaderboard forever. everything!\n\nAnd most importantly, share them! You might think your solution is similar to other participants, it might not, but it might not, and we want to each of them, to see if we can find a pattern. \n\nWe will also be circulating an exit survey later, stay tuned. ", "3033735": "Thank you for being a great host of an exciting competition @gordonyip ! Throughout the whole competition you answered all questions from the participants and were always clear about the competition setup. Especially with the distribution shift from train to test.\n\nCompetitions with synthetic datasets are historically difficult w.r.t. setup and keeping them leak free. And this one has been one of the best!", "3033766": "Thank you for the interesting competition!\nWe are currently working on summarizing the solution, with sum late subs to examin the effect of some of the components.\nThe full details should be available within the next few days at the latest!\n(As some of you may have noticed, we've been moving up the leaderboard quite rapidly over the past two days, and even we arenโt entirely sure yet what made our solution so strong!)" } }