LLM Assertiveness can be Mechanistically Decomposed into Emotional and Logical Components
This provides mechanistic evidence for the multi-component structure of LLM assertiveness, potentially helping mitigate overconfident behavior in high-stakes contexts.
The researchers investigated the internal basis of LLM overconfidence using mechanistic interpretability on Llama 3.2 models, finding that assertive representations decompose into emotional and logical components with distinct causal effects.
Large Language Models (LLMs) often display overconfidence, presenting information with unwarranted certainty in high-stakes contexts. We investigate the internal basis of this behavior via mechanistic interpretability. Using open-sourced Llama 3.2 models fine-tuned on human annotated assertiveness datasets, we extract residual activations across all layers, and compute similarity metrics to localize assertive representations. Our analysis identifies layers most sensitive to assertiveness contrasts and reveals that high-assertive representations decompose into two orthogonal sub-components of emotional and logical clusters-paralleling the dual-route Elaboration Likelihood Model in Psychology. Steering vectors derived from these sub-components show distinct causal effects: emotional vectors broadly influence prediction accuracy, while logical vectors exert more localized effects. These findings provide mechanistic evidence for the multi-component structure of LLM assertiveness and highlight avenues for mitigating overconfident behavior.