becca_test is the collection of test worlds for Becca, split out into a separate package for cleanliness and modularity. The version 0.8.x indicates that it's intended to be compatible with the latest Becca 0.8.x release.
The worlds in becca_test are toy worlds, but collectively they are quite challenging. On this current version, Becca tends to score around .55 out of 1 on a 10,000 time step run of the test suite. Although these worlds were built to test Becca, they can be used as an entry-level benchmark for any reinforcement learning algorithm that aspires to generality.