Logic-Regularized Verifier Elicits Reasoning from LLMs
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910196459110400 |
|---|---|
| author | Wang, Xinyu Sun, Changzhi Cheng, Lian Wu, Yuanbin Zhang, Dell Wang, Xiaoling Li, Xuelong |
| author_facet | Wang, Xinyu Sun, Changzhi Cheng, Lian Wu, Yuanbin Zhang, Dell Wang, Xiaoling Li, Xuelong |
| contents | Verifiers are crucial components for enhancing modern LLMs' reasoning capability. Typicalverifiers require resource-intensive superviseddataset construction, which is costly and faceslimitations in data diversity. In this paper, wepropose LOVER, an unsupervised verifier regularized by logical rules. LOVER treats theverifier as a binary latent variable, utilizinginternal activations and enforcing three logical constraints on multiple reasoning paths:negation consistency, intra-group consistency,and inter-group consistency (grouped by thefinal answer). By incorporating logical rulesas priors, LOVER can leverage unlabeled examples and is directly compatible with any offthe-shelf LLMs. Experiments on 10 datasetsdemonstrate that LOVER significantly outperforms unsupervised baselines, achieving performance comparable to the supervised verifier(reaching its 95% level on average). The sourcecode is publicly available at https://github.com/wangxinyufighting/llm-lover. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_05893 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Logic-Regularized Verifier Elicits Reasoning from LLMs Wang, Xinyu Sun, Changzhi Cheng, Lian Wu, Yuanbin Zhang, Dell Wang, Xiaoling Li, Xuelong Computation and Language Artificial Intelligence Verifiers are crucial components for enhancing modern LLMs' reasoning capability. Typicalverifiers require resource-intensive superviseddataset construction, which is costly and faceslimitations in data diversity. In this paper, wepropose LOVER, an unsupervised verifier regularized by logical rules. LOVER treats theverifier as a binary latent variable, utilizinginternal activations and enforcing three logical constraints on multiple reasoning paths:negation consistency, intra-group consistency,and inter-group consistency (grouped by thefinal answer). By incorporating logical rulesas priors, LOVER can leverage unlabeled examples and is directly compatible with any offthe-shelf LLMs. Experiments on 10 datasetsdemonstrate that LOVER significantly outperforms unsupervised baselines, achieving performance comparable to the supervised verifier(reaching its 95% level on average). The sourcecode is publicly available at https://github.com/wangxinyufighting/llm-lover. |
| title | Logic-Regularized Verifier Elicits Reasoning from LLMs |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2605.05893 |