Closed jvallikivi closed 1 month ago
Full list of mismatched data :
025d127b (training) 10fcaaa3 (training) 11852cab (training) 150deff5 (training) 1b60fb0c (training) 42a50994 (training) 469497ad (training) 6d0160f0 (training) 6d58a25d (training) 82819916 (training) 868de0fa (training) 9aec4887 (training) a9f96cdd (training) cbded52d (training) d511f180 (training) d687bc17 (training) dc433765 (training) e48d4e1a (training) ef135b50 (training) 12422b43 (evaluation) 17cae0c1 (evaluation) 310f3251 (evaluation) 423a55dc (evaluation) 48131b3c (evaluation) 54db823b (evaluation) 58e15b12 (evaluation) 7039b2d7 (evaluation) 8fbca751 (evaluation) a8610ef7 (evaluation) ad7e01d0 (evaluation) b0f4d537 (evaluation) b4a43f3b (evaluation) bd14c3bf (evaluation) c92b942c (evaluation) e7b06bea (evaluation) f8be4b64 (evaluation)
Note that I checked, and the data between Kaggle and the official ARC-AGI website is matching properly ! So it's just this repository that is outdated compared to the latest updates of the official data repository. @mxbi
Hey both, thanks a lot for pointing this out. The data was updated for arc-agi. Will publish a fix in the next few days
Hey, this has now been fixed in v0.1.0, which you can obtain with a pip install -U arckit
. Thanks!
I found a mismatch between a datapoint from the two data sources manually – there are potentially more mismatches.
Please see the output of the second training point
10fcaaa3
. Value in arc1.json:[[0, 0, 6, 0, 0, 0, 6, 0], [8, 8, 8, 8, 8, 8, 8, 8], [0, 6, 0, 8, 0, 6, 0, 8], [8, 0, 6, 0, 8, 0, 6, 0], [8, 8, 8, 8, 8, 8, 8, 0], [0, 6, 0, 0, 0, 6, 0, 0]]
Value in arc-agi_training_challenges.json (from Kaggle):[[0, 0, 6, 0, 0, 0, 6, 0], [8, 8, 8, 8, 8, 8, 8, 8], [0, 6, 0, 8, 0, 6, 0, 8], [8, 0, 6, 0, 8, 0, 6, 0], [8, 8, 8, 8, 8, 8, 8, 8], [0, 6, 0, 0, 0, 6, 0, 0]]
In order not to cause any issues, it might be good to either update
arc1.json
with the data from Kaggle and/or to create a function to easily import data downloaded from Kaggle.