2026· Annual Meeting of the Association for Computational Linguistics· pp. 19113-19126· 0 citations· 29 references
Computer Science
TL;DR
ODL-TempLLM leverages ontology learning to explicitly construct structured temporal knowledge, employs a symbolic reasoner to deductively reason about temporal relations and uses logic-constrained retrieval augmentation to obtain relevant facts.
Abstract
Temporal reasoning is crucial for large language models (LLMs) to understand event concurrency and complex temporal interactions in natural language. Recent approaches rely on the LLM to infer temporal relations between events and largely overlook the inherent structural nature of temporal relationships. In this work, we propose ODL-TempLLM ( O ntology-Guided and D escription L ogic–Constrained Temp oral Reasoning with LLM s), a novel paradigm for temporal reasoning with LLMs that shifts focus from internal inference to the explicit modeling of temporal structure. ODL-TempLLM leverages ontology learning to explicitly construct structured temporal knowledge, employs a symbolic reasoner to deductively reason about temporal relations and uses logic-constrained retrieval augmentation to obtain relevant facts. Experiments re-sults evaluated across three datasets via various LLM backbones show that our method outperforms state-of-the-art methods by 2.07–31.83 F1 points and 1.00–30.73 EM points, exhibiting strong generalization and highlighting the potential of explicit temporal reasoning.
Requirements engineering plays a central role in mechanical design, yet technical requirements remain predominantly expressed in natural language, limiting traceability, validation, and computational reasoning. This work presents an ontology-constrained pipeline for transforming natural-language engineering requirements into Industrial Ontologies Foundry (IOF)-grounded knowledge graphs enriched with QUDT-based quantitative semantics. The pipeline decomposes text blocks into individual prescriptive clauses, extracts structural slots and constraint atoms through a typed intermediate representation, normalizes quantitative expressions via QUDT unit and quantity-kind grounding, and instantiates IOF-compliant OWL ABox graphs. The transformation is implemented as a hybrid neuro-symbolic workflow that combines Large Language Models (LLMs) with typed intermediate representations, rule-based post-processing, and description-logic reasoning. Evaluation on a Formula SAE (FSAE) rules corpus, intentionally selected to stress quantitative constraint handling, shows good structural reliability in the evaluated setting. Slot-level extraction achieved a macro accuracy of 94.50%, while quantitative constraint identification reached 97.64% precision and 96.88% recall. Normalization coverage was 98.78%, with residual errors primarily attributable to quantity-kind disambiguation. At the graph level, 93.33% of grounded artifacts passed all ontology-conformance checks, with residual violations concentrated in requirement–specification linkage and specification typing rather than in quantitative-value modeling or systematic misuse of the IOF backbone. These results indicate that ontology-constrained LLM pipelines can support the formalization of engineering requirements into semantically explicit graph representations that are suitable for downstream querying, validation, and analysis.
A. Stefanone, M. Rossoni, Giorgio Colombo· Journal of Mechanical Design· 0 citations
SodaMem is presented, an evidence-grounded temporal graph memory that extracts typed FactEvents with mandatory provenance spans, persists mention time, occurrence time, and validity with SUPERSEDES/CONTRADICTS/UPDATES edges under hybrid lexical-dense indexing and answers via a planner-reader loop that gathers citable evidence before composing a final response.
MulTR, a comprehensive benchmark designed for multi-answer temporal reasoning from long unstructured contexts, is introduced, designed to systematically decouple temporal reasoning capabilities from the uncertainty of the number of answers.
Shaojuan Wu· Annual International ACM SIG...· 0 citations
OaK is presented, an ontology-as-a-kernel framework that dynamically constructs and refines task-oriented ontologies for LLM agents and shows that OaK improves standard LLM agents, strengthens evidence grounding, and boosts the reliability of multi-step reasoning.
Xiaohui Zhang, Zequn Sun, Cheng Yang et al.· 0 citations