Causally Grounded Mechanistic Interpretability for LLMs with Faithful Natural-Language Explanations

Researchers developed a pipeline to translate AI model internal mechanisms into human-understandable explanations, testing on GPT-2 Small. The study identified six attention heads responsible for 61.4% of model performance on a specific task, with LLM-generated explanations outperforming template-based approaches by 64%.

• →New pipeline bridges the gap between AI circuit-level analysis and natural language explanations for better interpretability.
• →Six attention heads in GPT-2 Small account for 61.4% of performance on the Indirect Object Identification task.
• →Circuit-based explanations achieved 100% sufficiency but only 22% comprehensiveness, revealing distributed backup mechanisms.
• →LLM-generated explanations outperformed template baselines by 64% on quality metrics.
• →No correlation found between model confidence and explanation faithfulness, with three identified failure categories.