The Effect of Model Size on LLM Post-hoc Explainability via LIME
This addresses the problem of explainability degradation in larger LLMs for researchers and practitioners, though it is incremental as it focuses on specific models and metrics.
The study investigated how model size affects LIME explainability for DeBERTaV3 models on NLI and ZSC tasks, finding that increased size does not improve plausibility despite better performance, indicating a misalignment with internal processes.
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.