Tag: mechanistic-interpretability

All the articles with the tag "mechanistic-interpretability".

• Testing Accessibility Knowledge Across Pythia Model Sizes

  18 Jan, 2026

  While using TransformerLens to explore how different model sizes handle digital accessibility concepts, an interesting pattern began to emerge. I started with GPT-2 and noticed that it did not understand the acronym, “WCAG.” Each time the question was run the model responded with something different that was either incorrect, had no meaning, or both.