Work

Polarity Composition Hypothesis: Causal Evidence from Negation in NLI

Year: 2026
Type: preprint
Abstract: We investigate how language models process negation in natural language inference (NLI). We propose the \textit{Polarity Composition Hypothesis}, a high-level causal model of this computation. Behavioral evaluations show that language-model predictions largely conform to the proposed hypothesis. Using Distributed Alignment Search, we identify causally effective subspaces aligned with the variables in the polarity-composition hypothesis. These representations transfer across negation forms and partially to a subset of MNLI. We release the dataset and code.
Institution: RIKEN
Language: English
Open Access status: green
OpenAlex ID: W7236262709
Cites: 23
Cited by:
Related to: 10