Wired for Overconfidence: A Mechanistic Perspective on Inflated Verbalized Confidence in LLMs
This addresses the issue of misleading users and unreliable uncertainty signals in LLMs, though it is incremental as it builds on existing mechanistic analysis methods.
The study tackled the problem of large language models being confidently wrong by analyzing the internal mechanisms that cause inflated verbalized confidence, finding that a compact set of circuits in middle-to-late layers drives this behavior and that targeted interventions can substantially improve calibration.
Large language models are often not just wrong, but \emph{confidently wrong}: when they produce factually incorrect answers, they tend to verbalize overly high confidence rather than signal uncertainty. Such verbalized overconfidence can mislead users and weaken confidence scores as a reliable uncertainty signal, yet its internal mechanisms remain poorly understood. We present a circuit-level mechanistic analysis of this inflated verbalized confidence in LLMs, organized around three axes: capturing verbalized confidence as a differentiable internal signal, identifying the circuits that causally inflate it, and leveraging these insights for targeted inference-time recalibration. Across two instruction-tuned LLMs on three datasets, we find that a compact set of MLP blocks and attention heads, concentrated in middle-to-late layers, consistently writes the confidence-inflation signal at the final token position. We further show that targeted inference-time interventions on these circuits substantially improve calibration. Together, our results suggest that verbalized overconfidence in LLMs is driven by identifiable internal circuits and can be mitigated through targeted intervention.