The Effect of Model Size on LLM Post-hoc Explainability via LIME

Henning Heyen, Amy Widdicombe, Noah Yamamoto Siegel, Philip Colin Treleaven, Maria Perez-Ortiz

Keywords: LLM Explainability, Large Language Models, LIME, Transformers, Alignment, Faithfulness, Plausibility, Scaling Laws

TL;DR: Novel insights on how model size in LLMs affects the quality of LIME explanations

Abstract: Large language models (LLMs) are becoming bigger to boost performance. However, little is known about how explainability is affected by this trend. This work explores LIME explanations for DeBERTaV3 models of four different sizes on natural language inference (NLI) and zero-shot classification (ZSC) tasks. We evaluate the explanations based on their faithfulness to the models’ internal decision processes and their plausibility, i.e. their agreement with human explanations. The key finding is that increased model size does not correlate with plausibility despite improved model performance, suggesting a misalignment between the LIME explanations and the models’ internal processes as model size increases. Our results further suggest limitations regarding faithfulness metrics in NLI contexts.

Submission Number: 98