Learning to decide which reasoning paths to take enables us to consider significantly deeper reasoning tasks and achieve SOTA results on CLUTRR, a dataset for testing systematic generalisation abilities of neural models -- Code and data at https://t.co/FWmC4SJKp6 2/N pic.twitter.com/B3Dvmewl5E— Pasquale Minervini (@PMinervini) July 14, 2020
No comments:
Post a Comment