Improving Model Robustness for Natural Language Inference

Add to your list(s) Download to your calendar using vCal

• Joe Stacey (Imperial College London)
• Friday 28 April 2023, 12:00-13:00
• Computer Lab, FW26.

If you have a question about this talk, please contact Michael Schlichtkrull.

Abstract:

Natural Language Inference (NLI) models are known to learn from biases within their training data, impacting how well the models generalise to other unseen datasets. Most methods to improve model robustness focus on preventing models learning from these biases, which can result in restrictive models and lower performance. We explore a range of alternative techniques to improve model robustness, including training models with human explanations, introducing a new logical reasoning framework, and generating domain-targeted data using GPT3 . We measure robustness by training models on SNLI and testing performance on MNLI , a challenging robustness setting where most prior work shows limited improvements.

Bio:

Joe is a 3rd year PhD student at Imperial College London supervised by Marek Rei. His research focuses on creating more robust NLP models that generalise better to unseen, out-of-distribution datasets. Joe is a recipient of the 2023 Apple Scholars in AI/ML PhD fellowship.

This talk is part of the NLIP Seminar Series series.

This talk is included in these lists:

• All Talks (aka the CURE list)
• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge Forum of Science and Humanities
• Cambridge Language Sciences
• Cambridge talks
• Chris Davis' list
• Computer Lab, FW26
• Department of Computer Science and Technology talks and seminars
• Graduate-Seminars
• Guy Emerson's list
• Interested Talks
• Language Sciences for Graduate Students
• NLIP Seminar Series
• PMRFPS's
• Simon Baker's List
• Trust & Technology Initiative - interesting events
• bld31
• ndk22's list
• ob366-ai4er
• rp587

Note that ex-directory lists are not shown.