4 papers
Automated grading of Linux/bash examinations using large language models: a four-level cognitive taxonomy approach
Manuel Alonso-Carracedo, Ruben Fernandez-Boullon, Pedro Celard +2
Scalable and reliable grading of command-line examinations remains a challenge in computing education, where rising enrolments make manual marking difficult and rule-based autograd…
CogTax: A Four-Level Cognitive Taxonomy for Command-Line Computing Education
Manuel Alonso-Carracedo, Ruben Fernandez-Boullon, Pedro Celard +2
As computing education expands beyond traditional programming into operational domains such as systems administration and command-line environments, existing pedagogical frameworks…
From Token Lists to Graph Motifs: Weisfeiler-Lehman Analysis of Sparse Autoencoder Features
Ruben Fernandez-Boullon, Pablo Magariños-Docampo, Javier Perez-Robles
Sparse autoencoders (SAEs) have become central to mechanistic interpretability, decomposing transformer activations into monosemantic features. Yet existing analyses characterise f…
Patch-Effect Graph Kernels for LLM Interpretability
Ruben Fernandez-Boullon, David N. Olivieri
Mechanistic interpretability aims to reverse-engineer transformer computations by identifying causal circuits through activation patching. However, scaling these interventions acro…