Language Modeling with Latent Situations

TL;DR

This paper introduces SituationSupervision, a method to enhance language model coherence by training them to explicitly represent and infer entity states, requiring minimal annotations for significant improvements.

Contribution

It proposes a novel approach combining auxiliary state modeling and latent state inference to improve language model coherence with limited supervision.

Findings

Achieves 4-11% coherence improvement.

Requires only a small amount of state annotations.

Applicable to both fine-tuning and prompting methods.

Abstract

Language models (LMs) often generate incoherent outputs: they refer to events and entity states that are incompatible with the state of the world described in their inputs. We introduce SituationSupervision, a family of approaches for improving coherence in LMs by training them to construct and condition on explicit representations of entities and their states. SituationSupervision has two components: an auxiliary situation modeling task that trains models to predict state representations in context, and a latent state inference procedure that imputes these states from partially annotated training data. SituationSupervision can be applied to both fine-tuning (by supervising LMs to encode state variables in their hidden representations) and prompting (by inducing LMs to interleave textual descriptions of entity states with output text). In both cases, SituationSupervision requires only a small number of state annotations to produce major coherence improvements (between 4-11%), showing that standard LMs can be sample-efficiently trained to model not just language but the situations it describes.