{"slug": "can-large-language-models-generalize-analogy-solving-like-children-can", "title": "Can Large Language Models Generalize Analogy Solving Like Children Can?", "summary": "A study published in Transactions of the Association for Computational Linguistics, Volume 14, pages 612–626, by Claire E. Stevenson, Alexandra Pafford, Han L. J. van der Maas, and Melanie Mitchell found that large language models failed to generalize letter-string analogy solving to unfamiliar domains where children and adults succeeded. Children and adults easily transferred their knowledge from Latin-alphabet analogies such as \"a b : a c :: j k : ?\" to a near transfer domain (Greek alphabet) and a far transfer domain (list of symbols), whereas the LLMs did not. The authors conclude this difference is evidence that these LLMs still struggle with robust human-like analogical transfer.", "body_md": "##### Abstract\n\nIn people, the ability to solve analogies such as “body: feet:: table: ?” emerges in childhood, and appears to transfer easily to other domains, such as the visual domain “(: ) :: < : ?”. Recent research shows that large language models (LLMs) can solve various forms of analogies. However, can LLMs generalize analogy solving to other domains like people can? To investigate this, we had children, adults, and LLMs solve a series of letter-string analogies (e.g., a b : a c :: j k : ?) in the Latin alphabet, in a near transfer domain (Greek alphabet), and a far transfer domain (list of symbols). Children and adults easily generalized their knowledge to unfamiliar domains, whereas LLMs did not. This key difference between human and AI performance is evidence that these LLMs still struggle with robust human-like analogical transfer.\n- Anthology ID:\n- 2026.tacl-1.28\n- Volume:\n- [Transactions of the Association for Computational Linguistics, Volume 14](https://aclanthology.org/volumes/2026.tacl-1/)\n- Month:\n- Year:\n- 2026\n- Address:\n- Cambridge, MA\n- Venue:\n- [TACL](https://aclanthology.org/venues/tacl/)\n- SIG:\n- Publisher:\n- MIT Press\n- Note:\n- Pages:\n- 612–626\n- Language:\n- URL:\n- [https://aclanthology.org/2026.tacl-1.28/](https://aclanthology.org/2026.tacl-1.28/)\n- DOI:\n- [10.1162/tacl.a.614](https://doi.org/10.1162/tacl.a.614)\n- Cite (ACL):\n- Claire E. Stevenson, Alexandra Pafford, Han L. J. van der Maas, and Melanie Mitchell. 2026. [Can Large Language Models Generalize Analogy Solving Like Children Can?](https://aclanthology.org/2026.tacl-1.28/) .*Transactions of the Association for Computational Linguistics* , 14:612–626.\n- Cite (Informal):\n- [Can Large Language Models Generalize Analogy Solving Like Children Can?](https://aclanthology.org/2026.tacl-1.28/) (Stevenson et al., TACL 2026)\n- PDF:\n- [https://aclanthology.org/2026.tacl-1.28.pdf](https://aclanthology.org/2026.tacl-1.28.pdf)", "url": "https://wpnews.pro/news/can-large-language-models-generalize-analogy-solving-like-children-can", "canonical_source": "https://aclanthology.org/2026.tacl-1.28/", "published_at": "2026-10-07 00:00:00+00:00", "updated_at": "2026-10-08 00:18:18.201186+00:00", "lang": "en", "topics": ["large-language-models", "ai-research", "machine-learning", "artificial-intelligence"], "entities": ["Claire E. Stevenson", "Alexandra Pafford", "Han L. J. van der Maas", "Melanie Mitchell", "Transactions of the Association for Computational Linguistics", "MIT Press"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/can-large-language-models-generalize-analogy-solving-like-children-can", "markdown": "https://wpnews.pro/news/can-large-language-models-generalize-analogy-solving-like-children-can.md", "text": "https://wpnews.pro/news/can-large-language-models-generalize-analogy-solving-like-children-can.txt", "jsonld": "https://wpnews.pro/news/can-large-language-models-generalize-analogy-solving-like-children-can.jsonld"}}