Papers by Manuel Vargas Guzmán
Teaching Small Language Models to Learn Logic through Meta-Learning (2026.eacl-long)
Copied to clipboard
| Challenge: | Large language models are increasingly evaluated on reasoning tasks, yet their logical abilities remain contested. |
| Approach: | They propose to apply few-shot meta-learning to large language models' reasoning domain to enable them to acquire abstract inference patterns that generalize to novel structures. |
| Outcome: | The proposed model outperforms GPT-4o and o3-mini on a syllogistic reasoning task. |
Testing the limits of logical reasoning in neural and hybrid models (2024.findings-naacl)
Copied to clipboard
| Challenge: | despite the successes of deep learning models, we still need to know more about how and what they learn. |
| Approach: | They create tests to analyze logical reasoning patterns in neural and hybrid models . they find that models can generalize logical thinking only to a limited degree . |
| Outcome: | The proposed models can capture elementary aspects of meaning but only to limited extent . authors say they need to understand how and what they learn . |
Prepositions Matter in Quantifier Scope Disambiguation (2022.coling-1)
Copied to clipboard
Aleksander Leczkowski, Justyna Grudzińska, Manuel Vargas Guzmán, Aleksander Wawer, Aleksandra Siemieniuk
| Challenge: | Existing work on how to integrate world knowledge into a QSD model has been limited . |
| Approach: | They use a scope-disambiguated corpus annotated with prepositional senses to integrate our knowledge into a machine learning model. |
| Outcome: | The proposed model is based on a scope-disambiguated corpus annotated with prepositional senses . Statistical analysis shows that prepositions have a positive impact on the learnability of automatic QSD systems. |