Methods for detecting forensics and attributing finetuned language models black-box access
2022, rewritten 2023
Proposal for a unified and standardised benchmark for dangerous capabilities in AI models
Non-trivial Fellowship, 2023
Researching situational awareness capabilities of gpt3.5 and gpt4 across different languages
Model evalulations hackathon, 2023
Using convolutional neural networks and GradientCam to detect and areas of interest dermatological cancer, Alzheimer’s, and Covid-19
2021