← Search

Diego Garcia-Olano

2 accepted papers

2026

Hallucination Reduction with CASAL: Contrastive Activation Steering for Amortized Learning

ICLR 2026poster

Large Language Models (LLMs) exhibit impressive capabilities but often hallucinate, confidently providing incorrect answers instead of admitting ignorance. Prior work has shown that models encode linear representations of their own knowledge and that activation steering can reduce hallucinations. Th…

Cited by 0SourceScholar
2023

Error Discovery By Clustering Influence Embeddings

NeurIPS 2023poster

We present a method for identifying groups of test examples---slices---on which a model under-performs, a task now known as slice discovery. We formalize coherence---a requirement that erroneous predictions, within a slice, should be wrong for the same reason---as a key property that any slice disco…

Cited by 4SourcePDFScholar