World model inspired sarcasm reasoning with large language model agents
Keito Inoshita, Shinnosuke Mizuno
TL;DR
This work reframes sarcasm understanding as a world-model inspired reasoning task and introduces WM-SAR, a modular framework that assigns five parallel LLM-based agents to compute literal meaning, context, normative expectation, inconsistency, and speaker intention. The agents' scalar outputs are deterministically combined by a lightweight logistic regression arbiter, enabling interpretable, tractable decision-making while leveraging the reasoning strengths of LLMs. Across three sarcasm benchmarks, WM-SAR outperforms deep-learning baselines and GPT-based prompting approaches, with ablations confirming the essential roles of semantic inconsistency and intention reasoning. The approach also demonstrates robustness to LLM backbone size, provides detailed explainability through agent rationales and LR weights, and offers a practical balance between accuracy and computation, suggesting a generalizable direction for structuring pragmatic NLP tasks.
Abstract
Sarcasm understanding is a challenging problem in natural language processing, as it requires capturing the discrepancy between the surface meaning of an utterance and the speaker's intentions as well as the surrounding social context. Although recent advances in deep learning and Large Language Models (LLMs) have substantially improved performance, most existing approaches still rely on black-box predictions of a single model, making it difficult to structurally explain the cognitive factors underlying sarcasm. Moreover, while sarcasm often emerges as a mismatch between semantic evaluation and normative expectations or intentions, frameworks that explicitly decompose and model these components remain limited. In this work, we reformulate sarcasm understanding as a world model inspired reasoning process and propose World Model inspired SArcasm Reasoning (WM-SAR), which decomposes literal meaning, context, normative expectation, and intention into specialized LLM-based agents. The discrepancy between literal evaluation and normative expectation is explicitly quantified as a deterministic inconsistency score, and together with an intention score, these signals are integrated by a lightweight Logistic Regression model to infer the final sarcasm probability. This design leverages the reasoning capability of LLMs while maintaining an interpretable numerical decision structure. Experiments on representative sarcasm detection benchmarks show that WM-SAR consistently outperforms existing deep learning and LLM-based methods. Ablation studies and case analyses further demonstrate that integrating semantic inconsistency and intention reasoning is essential for effective sarcasm detection, achieving both strong performance and high interpretability.