This project contails the code for training heuristic and learned time travel models.
The model set up and the code is based on EntNet [1] (https://github.com/facebookarchive/MemNN/tree/master/EntNet-babi).
To train a model, run the following command
th main.lua
The code corrects the errors in second order questions of the ToMi dataset [2] (https://github.com/facebookresearch/ToMi). To run:
python main.py
*[1] Mikael Henaff, Jason Weston, Arthur Szlam, Antoine Bordes, and Yann LeCun, "Tracking the World State with Recurrent Entity Networks", arXiv:1612.03969 [cs.CL].
*[2] Le, Matthew and Boureau, Y-Lan and Nickel, Maximilian, "Revisiting the Evaluation of Theory of Mind through Question Answering (https://www.aclweb.org/anthology/D19-1598) *.