# NLI under the Microscope: What Atomic Hypothesis Decomposition Reveals

arXiv: https://arxiv.org/abs/2502.08080 · HTML: https://mlsanity.com/paper/2502.08080/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 51 references that are in the map.
- First posted on arXiv 12 Feb 2025.

## Builds on (most cited first, 30 of 30)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference](https://mlsanity.com/paper/1704.05426/) (2017)
- [FEVER: a large-scale dataset for Fact Extraction and VERification](https://mlsanity.com/paper/1803.05355/) (2018)
- [FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation](https://mlsanity.com/paper/2305.14251/) (2023)
- [Adversarial NLI: A New Benchmark for Natural Language Understanding](https://mlsanity.com/paper/1910.14599/) (2019)
- [Annotation Artifacts in Natural Language Inference Data](https://mlsanity.com/paper/1803.02324/) (2018)
- [NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models](https://mlsanity.com/paper/2405.17428/) (2024)
- [Learning the Difference that Makes a Difference with Counterfactually-Augmented Data](https://mlsanity.com/paper/1909.12434/) (2019)
- [Social Chemistry 101: Learning to Reason about Social and Moral Norms](https://mlsanity.com/paper/2011.00620/) (2020)
- [WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation](https://mlsanity.com/paper/2201.05955/) (2022)
- [Evaluating the Logical Reasoning Ability of ChatGPT and GPT-4](https://mlsanity.com/paper/2304.03439/) (2023)
- [WiCE: Real-World Entailment for Claims in Wikipedia](https://mlsanity.com/paper/2303.01432/) (2023)
- [Unsupervised Question Decomposition for Question Answering](https://mlsanity.com/paper/2002.09758/) (2020)
- [Complex Claim Verification with Evidence Retrieved in the Wild](https://mlsanity.com/paper/2305.11859/) (2023)
- [Are Natural Language Inference Models IMPPRESsive? Learning IMPlicature and PRESupposition](https://mlsanity.com/paper/2004.03066/) (2020)
- [A Closer Look at Claim Decomposition](https://mlsanity.com/paper/2403.11903/) (2024)
- [Two Failures of Self-Consistency in the Multi-Step Reasoning of LLMs](https://mlsanity.com/paper/2305.14279/) (2023)
- [Does Putting a Linguist in the Loop Improve NLU Data Collection?](https://mlsanity.com/paper/2104.07179/) (2021)
- [Asking It All: Generating Contextualized Questions for any Semantic Role](https://mlsanity.com/paper/2109.04832/) (2021)
- [QUDEVAL: The Evaluation of Questions Under Discussion Discourse Parsing](https://mlsanity.com/paper/2310.14520/) (2023)
- [Measuring the Inconsistency of Large Language Models in Preferential Ranking](https://mlsanity.com/paper/2410.08851/) (2024)
- [Logical Reasoning with Span-Level Predictions for Interpretable and Robust NLI Models](https://mlsanity.com/paper/2205.11432/) (2022)
- [Atomic Inference for NLI with Generated Facts as Atoms](https://mlsanity.com/paper/2305.13214/) (2023)
- [Revisiting text decomposition methods for NLI-based factuality scoring of summaries](https://mlsanity.com/paper/2211.16853/) (2022)
- [Pregnant Questions: The Importance of Pragmatic Awareness in Maternal Health Question Answering](https://mlsanity.com/paper/2311.09542/) (2023)
- [QASem Parsing: Text-to-text Modeling of QA-based Semantics](https://mlsanity.com/paper/2205.11413/) (2022)
- [Zero-Shot Fact-Checking with Semantic Triples and Knowledge Graphs](https://mlsanity.com/paper/2312.11785/) (2023)
- [Evaluating Paraphrastic Robustness in Textual Entailment Models](https://mlsanity.com/paper/2306.16722/) (2023)
- [How often are errors in natural language reasoning due to paraphrastic variability?](https://mlsanity.com/paper/2404.11717/) (2024)

## Built on by (most cited first, 11 of 11)

- [WikiVideo: Article Generation from Multiple Videos](https://mlsanity.com/paper/2504.00939/) (2025)
- [Always Tell Me The Odds: Fine-grained Conditional Probability Estimation](https://mlsanity.com/paper/2505.01595/) (2025)
- [Atomic-SNLI: Fine-Grained Natural Language Inference through Atomic Fact Decomposition](https://mlsanity.com/paper/2601.06528/) (2026)
- [Bonsai: Interpretable Tree-Adaptive Grounded Reasoning](https://mlsanity.com/paper/2504.03640/) (2025)
- [Atomic Information Flow: A Network Flow Model for Tool Attributions in RAG Systems](https://mlsanity.com/paper/2602.04912/) (2026)
- [MERGE: Minimal Expression-Replacement GEneralization Test for Natural Language Inference](https://mlsanity.com/paper/2510.24295/) (2025)
- [Reverse-engineering NLI: A study of the meta-inferential properties of Natural Language Inference](https://mlsanity.com/paper/2601.05170/) (2026)
- [Decompose-and-Formalise: Recursively Verifiable Natural Language Inference](https://mlsanity.com/paper/2601.19605/) (2026)
- [Can We Do Interpretable NLI with Graphs Based on Atomic Propositions?](https://mlsanity.com/paper/2609.16814/) (2026)
- [Atomic Visual Entailment: Enhancing Zero-Shot Vision-Language Reasoning through Atomic Fact Decomposition and Learned Selection](https://mlsanity.com/paper/2610.05630/) (2026)
- [Learning Proposition-Level Links from Sentence-Level Supervision for Neuro-Symbolic NLI](https://mlsanity.com/paper/2610.15158/) (2026)

## Abstract (arXiv)

Decomposition of text into atomic propositions is a flexible framework allowing for the closer inspection of input and output text. We use atomic decomposition of hypotheses in two natural language reasoning tasks, traditional NLI and defeasible NLI, to form atomic sub-problems, or granular inferences that models must weigh when solving the overall problem. These atomic sub-problems serve as a tool to further understand the structure of both NLI and defeasible reasoning, probe a model's consistency and understanding of different inferences, and measure the diversity of examples in benchmark datasets. Our results indicate that LLMs still struggle with logical consistency on atomic NLI and defeasible NLI sub-problems. Lastly, we identify critical atomic sub-problems of defeasible NLI examples, or those that most contribute to the overall label, and propose a method to measure the inferential consistency of a model, a metric designed to capture the degree to which a model makes consistently correct or incorrect predictions about the same fact under different contexts.