Susan Wei
Susan Wei
Home
Research
Grants
Experience
Group
Talks
Workshop Organisation
Teaching
CV
Light
Dark
Automatic
LLM judges are easier to talk down than up: a Bayesian item response analysis
Guoyang Zheng
,
Susan Wei
,
Sevvandi Kandanaarachchi
September 2026
OpenReview
Type
Conference paper
Publication
NeurIPS 2026 Workshop on Trust in AI Evaluation (TAE)
Related
Advances in Singular Learning Theory for BDL
Elicitation without Backpropagation: Steering Model Behavior by Optimizing the Latent Posterior
What does a Bayes-filtered transformer believe? A predictive Monte Carlo approach
Uncertainty Decomposition for Bayes-Filtered Transformers via Bayesian Predictive Inference
TabMGP: Martingale Posterior with TabPFN
Cite
×