Hi, I'm Meg!
I currently work on research infrastructure at a company that makes large language models. Before that, I worked in research, crypto trading & quant finance, and studied computer science & physical sciences at Oxford & Cambridge.
I love music and have spent many thousands of hours on it. I studied violin & singing growing up, and now also play the bass, piano, guitar & drums.
My research:
The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" [2023]: LLMs fail to represent facts in an order-invariant way
Towards Understanding Sycophancy in Language Models [2023]: LLMs trained with Reinforcement Learning from Human Feedback show sycophantic behavior which may be caused by human preferences
Taken out of context: On measuring situational awareness in LLMs [2023]: LLMs can successfully pass tests in-context having only seen descriptions of the tests in their training data
Forecasting Rare Language Model Behaviors [2025]: We introduce a method that forecasts risks in deployment across OOMs more queries than tested in evaluation
Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training [2024]: We find that deceptive behaviors in LLMs can resist standard safety training techniques
Many-shot Jailbreaking [2024]
Evaluating feature steering: A case study in mitigating social biases [2024]
Steering Llama 2 via Contrastive Activation Addition [2023]: We develop a technique to control language model outputs by adding steering vectors to activations during inference
If you want to hear me talk: