Bayes and LLMs. The security of LLMs

Purpose

Promote discussions about potential research topics around two issues: a) Bayes for LLMs and LLMs for Bayes, b) Security of LLMs

Reading suggestions

Bayes and Deep Learning

• Introduction in “Current Advances in Neural Networks”

Very Large Scale Bayesian Inference

• Machine Learning and the Future of Bayesian Computation
• Bayesian Deep Learning in the Age of Large Scale AI
• Prior knowledge elicitation: The past, present, and future

Bayes and LLMs and Transformers

• Pathologies in priors and inference for transformers
• The Matrix: A Bayesian learning model for LLMs
• Bayesian reward models for LLM alignment

LLMs for Bayes

• “Can Transformers do Bayesian Inference”
• LLMs to enhance Bayesian optimization

Security in LLMs

• On the Dangers of Stochastic Parrots…
• PromptBench: Towards Evaluating the Robustness of Large Language Models on Adversarial Prompts

Date

May 8th at 10.45. @ICMAT

Facilitator

David Ríos Insua (ICMAT-CSIC)

Discussion Topics

The potential of transformers in Bayesian inference, adversarial attacks to LLMs, Causality and LLMs, etc.

Session recording

You can watch the session here.

Slides

Available here.