MorphNLI: A Stepwise Approach to Natural Language Inference Using Text Morphing

2025-02-13

AI Summary

This paper addresses the limited generalization and interpretability of cross-domain Natural Language Inference (NLI). We propose a novel “text deformation” paradigm: a large language model controllably edits the premise sentence stepwise toward the hypothesis, while a pre-trained NLI model (e.g., RoBERTa) incrementally evaluates semantic relation shifts at each step; intermediate representations are fused via an incremental aggregation mechanism to produce the final prediction. Crucially, this reformulates NLI as a progressive text transformation process, where the editing trajectory inherently yields human-readable, fine-grained reasoning paths. On multiple cross-domain NLI benchmarks, our method achieves up to a 12.6% absolute accuracy gain over strong baselines and maintains robust out-of-distribution performance. It thus simultaneously advances both predictive accuracy and intrinsic interpretability—offering the first framework that grounds NLI decisions in transparent, stepwise linguistic transformations.

Technology Category

• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

We introduce MorphNLI, a modular step-by-step approach to natural language inference (NLI). When classifying the premise-hypothesis pairs into {entailment, contradiction, neutral}, we use a language model to generate the necessary edits to incrementally transform (i.e., morph) the premise into the hypothesis. Then, using an off-the-shelf NLI model we track how the entailment progresses with these atomic changes, aggregating these intermediate labels into a final output. We demonstrate the advantages of our proposed method particularly in realistic cross-domain settings, where our method always outperforms strong baselines with improvements up to 12.6% (relative). Further, our proposed approach is explainable as the atomic edits can be used to understand the overall NLI label.

Problem

• Natural Language Inference classification
• Cross-domain performance improvement
• Explainable atomic edits tracking

Innovation

• Modular step-by-step NLI
• Text morphing with edits
• Explainable atomic changes tracking