Implicit Representations of Meaning in Neural Language Models
This addresses the problem of understanding the internal mechanisms of language models for researchers in NLP and AI, providing evidence for implicit meaning representation, though it is incremental in building on prior work.
The paper investigates whether neural language models like BART and T5 represent and reason about meaning beyond surface statistics, finding that they use dynamic representations of entities and situations similar to linguistic models, which support property readouts and affect language generation.
Does the effectiveness of neural language models derive entirely from accurate modeling of surface word co-occurrence statistics, or do these models represent and reason about the world they describe? In BART and T5 transformer language models, we identify contextual word representations that function as models of entities and situations as they evolve throughout a discourse. These neural representations have functional similarities to linguistic models of dynamic semantics: they support a linear readout of each entity's current properties and relations, and can be manipulated with predictable effects on language generation. Our results indicate that prediction in pretrained neural language models is supported, at least in part, by dynamic representations of meaning and implicit simulation of entity state, and that this behavior can be learned with only text as training data. Code and data are available at https://github.com/belindal/state-probes .