On the Calibration of Large Language Models and Alignment
This work addresses the reliability problem for users of large language models by providing insights into calibration throughout the training pipeline, though it is incremental as it builds on existing calibration methods.
The researchers systematically examined how pretraining and alignment training affect the calibration of large language models across generation, factuality, and understanding tasks, finding that training settings like parameter scales and data significantly influence calibration reliability.
As large language models attract increasing attention and find widespread application, concurrent challenges of reliability also arise at the same time. Confidence calibration, an effective analysis method for gauging the reliability of deep models, serves as a crucial tool for assessing and improving their reliability. However, such investigation has been comparatively underexplored. In this work, we conduct a systematic examination of the calibration of aligned language models throughout the entire construction process, including pretraining and alignment training. At each stage, we investigate how different training settings, such as parameter scales and training data, affect model calibration. To thoroughly assess model calibration, we evaluate models on three most concerned aspects: generation, factuality and understanding. Our work sheds light on whether popular LLMs are well-calibrated and how the training process influences model calibration.