Direct Confidence Alignment: Aligning Verbalized Confidence with Internal Confidence In Large Language Models
This addresses the need for more interpretable and trustworthy LLMs by improving confidence alignment, though it is incremental as it builds on existing calibration methods.
The paper tackles the problem of misalignment between large language models' internal confidence (from token probabilities) and verbalized confidence, which undermines calibration and trustworthiness, by proposing Direct Confidence Alignment (DCA) using Direct Preference Optimization to align these confidence measures; results show DCA improves alignment metrics on some model architectures but is ineffective on others.
Producing trustworthy and reliable Large Language Models (LLMs) has become increasingly important as their usage becomes more widespread. Calibration seeks to achieve this by improving the alignment between the model's confidence and the actual likelihood of its responses being correct or desirable. However, it has been observed that the internal confidence of a model, derived from token probabilities, is not well aligned with its verbalized confidence, leading to misleading results with different calibration methods. In this paper, we propose Direct Confidence Alignment (DCA), a method using Direct Preference Optimization to align an LLM's verbalized confidence with its internal confidence rather than ground-truth accuracy, enhancing model transparency and reliability by ensuring closer alignment between the two confidence measures. We evaluate DCA across multiple open-weight LLMs on a wide range of datasets. To further assess this alignment, we also introduce three new calibration error-based metrics. Our results show that DCA improves alignment metrics on certain model architectures, reducing inconsistencies in a model's confidence expression. However, we also show that it can be ineffective on others, highlighting the need for more model-aware approaches in the pursuit of more interpretable and trustworthy LLMs.