Do Large Language Models know who did what to whom?
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910919244644352 |
|---|---|
| author | Denning, Joseph M. Guo, Xiaohan Hannah Snefjella, Bryor Blank, Idan A. |
| author_facet | Denning, Joseph M. Guo, Xiaohan Hannah Snefjella, Bryor Blank, Idan A. |
| contents | Large Language Models (LLMs) are commonly criticized for not understanding language. However, many critiques focus on cognitive abilities that, in humans, are distinct from language processing. Here, we instead study a kind of understanding tightly linked to language: inferring who did what to whom (thematic roles) in a sentence. Does the central training objective of LLMs-word prediction-result in sentence representations that capture thematic roles? In two experiments, we characterized sentence representations in four LLMs. In contrast to human similarity judgments, in LLMs the overall representational similarity of sentence pairs reflected syntactic similarity but not whether their agent and patient assignments were identical vs. reversed. Furthermore, we found little evidence that thematic role information was available in any subset of hidden units. However, some attention heads robustly captured thematic roles, independently of syntax. Therefore, LLMs can extract thematic roles but, relative to humans, this information influences their representations more weakly. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_16884 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Do Large Language Models know who did what to whom? Denning, Joseph M. Guo, Xiaohan Hannah Snefjella, Bryor Blank, Idan A. Computation and Language Large Language Models (LLMs) are commonly criticized for not understanding language. However, many critiques focus on cognitive abilities that, in humans, are distinct from language processing. Here, we instead study a kind of understanding tightly linked to language: inferring who did what to whom (thematic roles) in a sentence. Does the central training objective of LLMs-word prediction-result in sentence representations that capture thematic roles? In two experiments, we characterized sentence representations in four LLMs. In contrast to human similarity judgments, in LLMs the overall representational similarity of sentence pairs reflected syntactic similarity but not whether their agent and patient assignments were identical vs. reversed. Furthermore, we found little evidence that thematic role information was available in any subset of hidden units. However, some attention heads robustly captured thematic roles, independently of syntax. Therefore, LLMs can extract thematic roles but, relative to humans, this information influences their representations more weakly. |
| title | Do Large Language Models know who did what to whom? |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2504.16884 |