Gardado en:
| Autor Principal: | |
|---|---|
| Formato: | Recurso digital |
| Idioma: | inglés |
| Publicado: |
Zenodo
2026
|
| Subjects: | |
| Acceso en liña: | https://doi.org/10.5281/zenodo.18148498 |
| Tags: |
Engadir etiqueta
Sen Etiquetas, Sexa o primeiro en etiquetar este rexistro!
|
Table of Contents:
- <p>Large language models frequently produce outputs that are syntactically fluent yet semantically invalid.<br>While recent verification approaches focus on reasoning chains or factual knowledge, many semantic failures occur without explicit reasoning or missing facts.<br>Such errors include temporal inconsistencies, numerical impossibilities, and semantic type clashes, which cannot be reliably detected by reasoning-based or knowledge-based methods.</p> <p>This paper introduces <strong>Catelingo</strong>, a constraint-based semantic validity verifier that detects these failures by checking the satisfiability of explicit semantic constraints.<br>Rather than evaluating truth, likelihood, or factual correctness, Catelingo defines semantic validity as constraint satisfiability induced by an input sentence.<br>The proposed approach requires neither reasoning chains nor knowledge retrieval, and does not rely on model retraining.</p> <p>We implement a toy version of Catelingo using a small, sense-level lexicon and explicit constraint propagation over syntactic dependency structures.<br>Experiments demonstrate that Catelingo correctly detects semantic <strong>no-go</strong> cases involving temporal ordering violations, numerical range violations, and semantic type incompatibilities.<br>We further show that metaphorical expressions can be selectively permitted through explicit degeneration rules, and that domain adaptation can be achieved by switching constraint profiles rather than retraining models.</p> <p>These results suggest that constraint-based semantic verification provides a lightweight and scalable complement to existing reasoning- and knowledge-based verification methods, addressing a class of semantic failures that they do not cover.<br>An open-source reference implementation and deterministic test cases are provided for reproducibility.<br>The source code for the verification, <strong>Catelingo</strong>, is open-sourced at <a title="Catelingo" href="https://github.com/ShinobuMiya/Catelingo">https://github.com/ShinobuMiya/Catelingo</a></p> <p>This work is intended as a design-oriented technical report and is not tied to a specific benchmark or leaderboard.</p>