Liam Bai Blog RSS Feed

LLM safeguards, the Swiss cheese model, persona hijacks, prefix injection, encoded prompts, many-shot attacks.

PD-1/PD-L1, immuno-oncology, checkpoint inhibition, binding visualizations, antibody anatomy.

The Minimum Description Length (MDL) principle, Kolmogorov complexity, linear regression, data compression and learning.

Applying the logit lens to ESM-2.

Predicting variant effects with variational autoencoders.

Predicting protein function using deep generative models.

Protein design by hallucination.

Learning protein representations.

Predicting protein structure and function.