1. Truth-value judgment in language models: belief directions are context sensitive
- Author
-
Schouten, Stefan F., Bloem, Peter, Markov, Ilia, and Vossen, Piek
- Subjects
Computer Science - Computation and Language - Abstract
Recent work has demonstrated that the latent spaces of large language models (LLMs) contain directions predictive of the truth of sentences. Multiple methods recover such directions and build probes that are described as getting at a model's "knowledge" or "beliefs". We investigate this phenomenon, looking closely at the impact of context on the probes. Our experiments establish where in the LLM the probe's predictions can be described as being conditional on the preceding (related) sentences. Specifically, we quantify the responsiveness of the probes to the presence of (negated) supporting and contradicting sentences, and score the probes on their consistency. We also perform a causal intervention experiment, investigating whether moving the representation of a premise along these belief directions influences the position of the hypothesis along that same direction. We find that the probes we test are generally context sensitive, but that contexts which should not affect the truth often still impact the probe outputs. Our experiments show that the type of errors depend on the layer, the (type of) model, and the kind of data. Finally, our results suggest that belief directions are (one of the) causal mediators in the inference process that incorporates in-context information.
- Published
- 2024