发表机构
Stanford University; University of Calgary（斯坦福大学; 卡尔加里大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对本地SLMs科学问答的约束，提出EGT-KG检索框架，经六维评估框架在生物聚合物结合土壤复合材料基准上验证，其多数场景优于普通RAG，llama3:8b变体提升显著。
AI 中文摘要
针对新兴科学研究领域，本地小型语言模型（SLMs）正变得更具吸引力，因为它们相比大型语言模型能提供更强的隐私控制和更稳定的部署流程。但在实际应用中，基于SLMs的科学问答常面临不可避免的约束：文献集合规模小、证据碎片化、上下文窗口有限以及推理能力不足。我们提出了证据基类型化知识图谱（EGT-KG），这是一种用于改进本地SLMs信息检索的检索框架。我们评估了三种问答设置：普通检索增强生成（RAG）工作流，以及两种EGT-KG工作流——自动生成关系模式（AS）和专家定义关系模式（ES）。我们在六维评估框架（S3CRF：合理性、正确性、完整性、简洁性、相关性、流畅性）下，基于生物聚合物结合土壤复合材料文献基准开展实验，结果显示EGT-KG在多数设置下优于普通RAG方法，其中llama3:8b的最佳改进为：AS/ES EGT-KG变体的最终得分分别为70.37（提升14.67%）和68.82（提升12.14%）。
英文摘要
For emerging scientific research domains, local Small Language Models (SLMs) are becoming more attractive, as they offer stronger privacy control and more stable deployment pipelines than Large Language Models. However, in practice, scientific question-answering on SLMs often operates under inevitable constraints: small literature collections, fragmented evidence, limited context window and reasoning abilities. We propose the Evidence-Grounded Typed Knowledge Graph (EGT-KG), a retrieval framework to improve information retrieval with local SLMs. We assessed three question-answering settings: a vanilla Retrieval-Augmented Generation (RAG) workflow and two EGT-KG workflows: an automatically generated relation schema (AS) and an expert-defined relation schema (ES). Our experiments were evaluated with a six-dimensional evaluation framework (S3CRF: Soundness, Correctness, Completeness, Conciseness, Relevance, Fluency) on a Biopolymer-bound Soil Composite literature benchmark, showing that EGT-KG outperforms the vanilla RAG method in most settings, with the best improvement from llama3:8b: a Final Score of 70.37 (+14.67%) and 68.82 (+12.14%) by AS/ES EGT-KG variants.
CommentsAccepted in EMNLP Industry track 2026
Journal refEMNLP Industry track 2026
