[Adapt] [Seminar] Probing Neural Network Comprehension of Natural Language Arguments

[廖千姿]

Hello Adapters,

BERT has caused a stir in the Machine Learning community by presenting state-of-the-art results in a wide variety of NLP tasks.
However, recently someone suspected that BERT's surprising performance on a particular dataset can be entirely accounted for in terms of exploiting spurious statistical cues in the data. Without them it is no better than random.
I would like to show you the analysis in details and the solution.

Time: Wed 4:30pm
Venue: SEIEE 4-414

See you there!

Best Regards,
Eve