Tag
A tweet recommending Nathan Cantafio's blog post 'MLE is not intuitive', which discusses common misconceptions about Maximum Likelihood Estimation in an accessible way.
This paper analyzes hallucination in large language models as a structural consequence of three architectural decisions: self-attention's co-occurrence learning, maximum likelihood estimation training objective, and autoregressive decoding's left-to-right commitment. It maps each mechanism to specific hallucination types and argues that dataset pathologies amplify but do not cause these vulnerabilities.
This article explains Behavioral Cloning, an imitation learning technique used to train policies from expert demonstrations, discussing its theoretical basis in maximum likelihood estimation and its historical applications like AlphaGo.