This is the repository to evaluate the question-answering module for physical reasoning in PHYRE benchmark.
Our goal is to let agent learn the generalized physical representation through reasoning module.
(Deep Learning, Korea University)
Learned embedding from question-answer module is applied into the main network through FiLM-based attention.
Models | Cross | Within |
---|---|---|
DQN | 36.8 | 77.6 |
ATQAN | 39.0 | 80.2 |
Our model was able to surpass the baseline through AUCESS score, in PHYRE-B task.