UTD-HLTRI at SemEval 2026 Task 4: Reasoning like an Expert for Inferring Narrative Similarity

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, short

Summary

UTD.HLTRISIM.NARRATIVES is a novel method developed by UTD-HLTRI for SemEval 2026 Task 4, focused on inferring narrative story similarity. This challenging problem necessitates reasoning across three key narrative aspects: the abstract theme, the course of action, and the outcomes. The proposed approach integrates contrastive reasoning prompting with a meticulous selection of few-shot examples to effectively guide a Large Language Model (LLM) in making comparative similarity decisions between narratives. Furthermore, the method incorporates a curriculum learning framework. This framework systematically orders narrative triplets presented to the LLM, utilizing a specialized score that quantifies the influence of common narrative elements and information derived from various distractors of narrative similarity.

Key takeaway

For NLP Engineers developing narrative understanding systems, this work suggests a robust approach to improve Large Language Model performance on similarity tasks. You should consider integrating contrastive reasoning prompting and a curriculum learning framework, as demonstrated by UTD.HLTRISIM.NARRATIVES. This method carefully selects few-shot examples and scores narrative aspects. It enhances an LLM's ability to discern subtle differences in themes, actions, and outcomes, leading to more accurate similarity judgments.

Key insights

The UTD.HLTRISIM.NARRATIVES method guides LLMs for narrative similarity using contrastive prompting and curriculum learning.

Principles

• Narrative similarity requires reasoning on theme, action, and outcomes.
• Curriculum learning can order examples for LLM training.
• Distractors of similarity can inform scoring.

Method

UTD.HLTRISIM.NARRATIVES combines contrastive reasoning prompting with few-shot example selection to guide an LLM. A curriculum learning framework orders narrative triplets using a score based on common aspects and distractors.

In practice

• Implement contrastive prompting for LLM narrative tasks.
• Design few-shot examples to guide LLM reasoning.
• Develop a scoring mechanism for curriculum learning.

Topics

• Narrative Similarity
• SemEval 2026
• Large Language Models
• Contrastive Prompting
• Curriculum Learning
• Few-shot Learning

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• Disentangling Linguistic Relatedness from Task Alignment in Cross-Lingual Transfer — Cross-lingual fine-tuning primarily improves task-format alignment, not language-specific knowledge transfer, especially for weaker models.
• Retell, Reward, Repeat: Reinforcement Learning for Narrative Theory-Informed Story Retelling — Reinforcement learning with narrative theory-informed AI feedback improves story diversity and human narrative alignment.
• Narration-of-Thought: Inference-Time Scaffolding for Defeasible Ethical Reasoning in Large Language Models — Narration-of-Thought (NoT) enhances LLM ethical reasoning by structuring chain-of-thought to explicitly address stakeholders and uncertainties.
• LMs as Task-Specific Knowledge Bases: An Interpretability Analysis — Language models encode factual knowledge in a task-specific manner, undermining the consistent "knowledge base" analogy.
• ttda704 at SemEval-2026 Task 4: Modeling Narrative Structures via Pseudonymization and Multi-View Sentence Alignment — Using contrastive learning with fine-tuned sentence transformers and multi-view alignment effectively models narrative structures and similarity.
• English is Not All You Need: Systematically Exploring the Role of Multilinguality in LLM Post-Training — Multilingual post-training significantly improves LLM performance across languages and tasks, outperforming English-only approaches.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.