[minimaxir]

Yes, just found it in the details: https://contest.openai.com/details

From the downloaded CSV of game states, all stages from the 3 games are valid training stages, and the custom stages used for testing are derived from all 3 games. Yikes.

I really hope those custom stages use the same art assets as the original games.

[oh_sigh]

Every test run could theoretically start with a jump or some other behavior to exercise the different physics or whatever among the 3 games, and use a discriminator network to feed into one of three networks individually trained on each game?

[tedivm]

The whole point of the OpenAI exercise is to improve transfer learning- a winning project should be able to work off of a single model.

For an example of a system that does learn new rules using a single model check out this post from Vicarious[0].

[0] https://www.vicarious.com/2017/08/07/general-game-playing-wi...