Atomic-SNLI: Fine-Grained Natural Language Inference through Atomic Fact Decomposition

TL;DR

Atomic-SNLI introduces a new dataset for fine-grained natural language inference at the atomic fact level, improving models' reasoning and explainability by decomposing hypotheses into individual facts.

Contribution

The paper presents Atomic-SNLI, a dataset that enables models to perform atomic-level inference, addressing the gap in fine-grained reasoning and interpretability in NLI.

Findings

Models trained on Atomic-SNLI show improved atomic reasoning.

Fine-tuning on Atomic-SNLI maintains strong sentence-level performance.

Atomic-SNLI enhances model explainability through fact-level inference.

Abstract

Current Natural Language Inference (NLI) systems primarily operate at the sentence level, providing black-box decisions that lack explanatory power. While atomic-level NLI offers a promising alternative by decomposing hypotheses into individual facts, we demonstrate that the conventional assumption that a hypothesis is entailed only when all its atomic facts are entailed fails in practice due to models' poor performance on fine-grained reasoning. Our analysis reveals that existing models perform substantially worse on atomic level inference compared to sentence level tasks. To address this limitation, we introduce Atomic-SNLI, a novel dataset constructed by decomposing SNLI and enriching it with carefully curated atomic level examples through linguistically informed generation strategies. Experimental results demonstrate that models fine-tuned on Atomic-SNLI achieve significant improvements in atomic reasoning capabilities while maintaining strong sentence level performance, enabling both accurate judgements and transparent, explainable results at the fact level.