projects

Interpretable Evaluation of Language Model Training Trajectories

Can we automatically discover what is learnt at various stages of training that influences model behaviour?

A Mechanistic Comparison of Representations of Deep RL Algorithms

What is learnt differently by RL algorithms that make them better or worse than others?

Protein Function Prediction

Kaggle Challenge

Adobe GenStudio - Content Intelligence

Providing and explaining marketing features from content

Adobe Dynamic Chat

Chat Summarisation Pipeline

Adobe Express

Integrated AI-based recommendations

Contextual Alchemy

LLMs to enhance readability of online content

Hierarchical Content Understanding

Extract hierarchical topics from content

Clarifying User Intent in Enterprise Search

providing clarifying suggestions for ambiguous queries

Analysing the Impact of COVID-19 on Food Security

Using data from daily covid-19 cases, demographic distribution, grocery store mobility, and household surveys to understand food security in the face of COVID-19 across different clusters of the population.

Interpretable QA for Privacy Policies

It is hard to read privacy policy documents. Not sure whether your searches are tracked? Why not use our QA system to find out?

News Media Bias Detection

A tool to provide context for online news consumption.

Detecting COVID-19 Onset from Wearables

Maybe your fitness trackers can tell you if you are infected with COVID-19

academic