Self-Aware Knowledge Probing: Evaluating Language Models' Relational Knowledge through Confidence Calibration
This work addresses the reliability of knowledge probing in language models, which is crucial for AI researchers and practitioners, though it is incremental in refining evaluation methods.
The paper tackled the problem of evaluating language models' relational knowledge by proposing a calibration probing framework that assesses model confidence across three modalities, revealing that most models are overconfident and even large models fail to accurately encode linguistic confidence semantics.
Knowledge probing quantifies how much relational knowledge a language model (LM) has acquired during pre-training. Existing knowledge probes evaluate model capabilities through metrics like prediction accuracy and precision. Such evaluations fail to account for the model's reliability, reflected in the calibration of its confidence scores. In this paper, we propose a novel calibration probing framework for relational knowledge, covering three modalities of model confidence: (1) intrinsic confidence, (2) structural consistency and (3) semantic grounding. Our extensive analysis of ten causal and six masked language models reveals that most models, especially those pre-trained with the masking objective, are overconfident. The best-calibrated scores come from confidence estimates that account for inconsistencies due to statement rephrasing. Moreover, even the largest pre-trained models fail to encode the semantics of linguistic confidence expressions accurately.