arXiv：cs.AI· Zhichao Hou, Ferhat Erata, Joe Lilien, MohamadAli Torkamani·· 4 小时前AI 评分34
分层一致性蒸馏：用微调提升自然语言到逻辑公式的翻译准确率
Stratified Consistency Distillation for Natural Language Formalization
AI 导读
研究提出分层一致性蒸馏方法，用前沿 LLM 为每个输入生成 K 个逻辑翻译并按语义等价聚类，再依据熵值分别采用多数投票、LLM-as-a-Judge 或统一/弃权策略筛选伪标签，微调更小的模型。实验显示 Pass@K 与新的 Equivalent Logical Similarity 指标均取得显著且一致的提升，表明一致性蒸馏可推进逻辑翻译。
正文
Abstract:Neurosymbolic reasoning has shown promising success in addressing complex reasoning tasks by combining large language models (LLMs) and symbolic solvers. While this approach shows promise, a fundamental challenge remains: improving the accuracy of translations from natural language to logical formulas. Current methods predominantly rely on prompt engineering, which is difficult to scale across different domains and input formats. Drawing inspiration from the success of fine-tuning in other model adaptation and alignment applications, we propose a fine-tuning-based Stratified Consistency Distillation approach: (1) We generate K logical translations per input using a frontier LLM and cluster them by semantic equivalence (2) Based on the entropy level, we apply majority voting (low entropy), LLM-as-a-Judge (medium entropy), or unification/abstention (high entropy), and (3) fine-tune a smaller model using the selected pseudo-labels. Our experiments show significant and consistent improvements in both Pass@K and our novel Equivalent Logical Similarity metrics, demonstrating the potential of advancing logical translation through consistency distillation.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.30258 [cs.CL] |
| (or arXiv:2608.30258v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2608.30258 arXiv-issued DOI via DataCite |
Submission history
From: Zhichao Hou [view email]
[v1]
Mon, 31 Aug 2026 05:08:08 UTC (7,413 KB)
[v2]
Thu, 1 Oct 2026 01:04:17 UTC (7,423 KB)
[v3]
Fri, 2 Oct 2026 17:23:52 UTC (7,416 KB)
来源：arXiv：cs.AI · arxiv.org