Relational Schemata in BERT Are Inducible, Not Emergent: A Study of Performance vs. Competence in Language Models
This addresses the problem of distinguishing true conceptual competence from surface-level associations in language models for AI researchers, showing that relational schemata are inducible rather than emergent, which is incremental in understanding model capabilities.
The study investigated whether BERT encodes abstract relational schemata by analyzing its internal representations for concept pairs across different relation types, finding that high classification accuracy indicates latent signals but relational organization only emerges after fine-tuning, not from pretraining alone.
While large language models like BERT demonstrate strong empirical performance on semantic tasks, whether this reflects true conceptual competence or surface-level statistical association remains unclear. I investigate whether BERT encodes abstract relational schemata by examining internal representations of concept pairs across taxonomic, mereological, and functional relations. I compare BERT's relational classification performance with representational structure in [CLS] token embeddings. Results reveal that pretrained BERT enables high classification accuracy, indicating latent relational signals. However, concept pairs organize by relation type in high-dimensional embedding space only after fine-tuning on supervised relation classification tasks. This indicates relational schemata are not emergent from pretraining alone but can be induced via task scaffolding. These findings demonstrate that behavioral performance does not necessarily imply structured conceptual understanding, though models can acquire inductive biases for grounded relational abstraction through appropriate training.