Calibrating LLMs with Information-Theoretic Evidential Deep Learning
This work addresses the problem of overconfidence in LLMs for users requiring high levels of confidence calibration, such as those in high-stakes domains, by providing a more trustworthy and reliable approach.
The authors tackled the problem of overconfidence in fine-tuned large language models, achieving improved calibration and uncertainty estimates through their proposed IB-EDL approach, which outperformed existing methods. IB-EDL demonstrated its effectiveness across various fine-tuned LLMs and tasks.
Fine-tuned large language models (LLMs) often exhibit overconfidence, particularly when trained on small datasets, resulting in poor calibration and inaccurate uncertainty estimates. Evidential Deep Learning (EDL), an uncertainty-aware approach, enables uncertainty estimation in a single forward pass, making it a promising method for calibrating fine-tuned LLMs. However, despite its computational efficiency, EDL is prone to overfitting, as its training objective can result in overly concentrated probability distributions. To mitigate this, we propose regularizing EDL by incorporating an information bottleneck (IB). Our approach IB-EDL suppresses spurious information in the evidence generated by the model and encourages truly predictive information to influence both the predictions and uncertainty estimates. Extensive experiments across various fine-tuned LLMs and tasks demonstrate that IB-EDL outperforms both existing EDL and non-EDL approaches. By improving the trustworthiness of LLMs, IB-EDL facilitates their broader adoption in domains requiring high levels of confidence calibration. Code is available at https://github.com/sandylaker/ib-edl.