Start by reviewing the open Issues list, in particular, Issue #3 Scoring Criteria.
See the growing list of Accessibilty Tests in WPT.
As of June 28th, we've completed 70% of our scoring criteria, including: reaching consensus on testing approach, adding testing tools to wpt to test computed role and label, adding a webdriver api for testing computed role, and landing that in Chrome, Webkit, and Gecko.
As of June 28th, we are part way through our remaining tasks (p1 and p2 tests) using the new infrastructure, which adds an additional ~8% to our score on a prorated basis.
We are tracking this progress in issue #3.