University of Cambridge > Talks.cam > Language Technology Lab Seminars > Towards Knowledge-Robust and Multimodally-Grounded NLP

Towards Knowledge-Robust and Multimodally-Grounded NLP

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

If you have a question about this talk, please contact Marinela Parovic.

In this talk, I will present our group’s work on NLP models that are knowledge-robust and multimodally-grounded. First, we will describe multi-task and reinforcement learning methods to incorporate novel auxiliary-skill tasks such as saliency, entailment, and back-translation validity (including bandit-based methods for automatic auxiliary task selection+mixing and multi-reward mixing). Next, we will discuss developing adversarial robustness against reasoning shortcuts, missing commonsense gaps, and cross-domain/lingual generalization in QA and dialogue models (including auto-adversary generation). Lastly, we will discuss multimodally-grounded models which condition and reason on dynamic spatio-temporal information in images and videos, and action-based robotic navigation and assembling tasks (including commonsense reasoning for ambiguous robotic instructions).

This talk is part of the Language Technology Lab Seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2024 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity