Mistake in the evaluation

A lot of models are outputing and working till moves 1,2,3,4,5 and then stop working. Then you are calculating the ACPL for only those moves whcih is wrong. If a model fails to complete a game that game should be disregarded

4 Likes

I have the same problem. Hopefully this will be fixed soon. :smiling_face_with_tear:

1 Like

Hi Rakshit, we have identified the issue and are working on a resolution. Thank you for flagging!

2 Likes

Hi Alexa,

I still see no steps taken, the issue is a major one and is affecting the leaderboard. If the AI model fails to finish the game then you shouldn’t take it into consideration as simple as that. Make sure to update it and rerun the submissions

The entire hackathon feels extremley disorganized and there are a lot of unanswered questions that they refuse to answer or dont acknowledge