Capturing Symmetry and Antisymmetry in Language Models through Symmetry-Aware Training Objectives
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913804475957248 |
|---|---|
| author | Yuan, Zhangdie Vlachos, Andreas |
| author_facet | Yuan, Zhangdie Vlachos, Andreas |
| contents | Capturing symmetric (e.g., country borders another country) and antisymmetric (e.g., parent_of) relations is crucial for a variety of applications. This paper tackles this challenge by introducing a novel Wikidata-derived natural language inference dataset designed to evaluate large language models (LLMs). Our findings reveal that LLMs perform comparably to random chance on this benchmark, highlighting a gap in relational understanding. To address this, we explore encoder retraining via contrastive learning with k-nearest neighbors. The retrained encoder matches the performance of fine-tuned classification heads while offering additional benefits, including greater efficiency in few-shot learning and improved mitigation of catastrophic forgetting. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_16312 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Capturing Symmetry and Antisymmetry in Language Models through Symmetry-Aware Training Objectives Yuan, Zhangdie Vlachos, Andreas Computation and Language Capturing symmetric (e.g., country borders another country) and antisymmetric (e.g., parent_of) relations is crucial for a variety of applications. This paper tackles this challenge by introducing a novel Wikidata-derived natural language inference dataset designed to evaluate large language models (LLMs). Our findings reveal that LLMs perform comparably to random chance on this benchmark, highlighting a gap in relational understanding. To address this, we explore encoder retraining via contrastive learning with k-nearest neighbors. The retrained encoder matches the performance of fine-tuned classification heads while offering additional benefits, including greater efficiency in few-shot learning and improved mitigation of catastrophic forgetting. |
| title | Capturing Symmetry and Antisymmetry in Language Models through Symmetry-Aware Training Objectives |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2504.16312 |