SynEL:对于连接实体的合成基准
Ilia Karpov1, Alexander Kirillovich2, Elisaveta Goncharova1
1HSE University, Moscow, Russia.
PloS one
|January 8, 2026
概括
大型语言模型 (LLM) 由于缺乏基准,难以提取特定领域的知识. 一个新的基准,SynEL和合成数据提高了LLM在这些任务上的表现.
科学领域:
- 人工智能的人工智能
- 自然语言处理自然语言处理.
- 知识表示 知识表示
背景情况:
- 大型语言模型 (LLM) 显示出从各种领域的文本构建知识图的前景.
- 法学士的表现受到特定领域语言的阻碍,需要专门的基准来完成知识图构建任务.
- 在知识图表构建中评估LLM的现有领域特定基准是有限的.
研究的目的:
- 引入SynEL,一个用于评估基于文本的知识提取方法的新基准.
- 为解决LLM评估领域特定基准的稀缺问题.
- 提出和验证生成合成数据集的方法,以提高LLM的性能.
主要方法:
- 开发了一个全面的方法来构建SynEL基准.
- 为生成合成数据集提出了两种不同的方法.
- 通过使用SynEL基准和合成数据,对知识提取任务的LLM绩效进行评估.
- 在LLM输出中评估累积幻觉.
主要成果:
- 与高资源实体相比,当前的LLM在提取低资源实体时,表现显著下降 (在微型F1得分中高达25个绝对点).
- 将合成数据集纳入培训过程导致微型F1得分提高了高达10个绝对点.
- 通过客户支持对话来验证SynEL基准.
结论:
- 像SynEL这样的专业基准对于准确评估和优化LLM对于特定领域的知识图表构建至关重要.
- 合成数据生成提供了一个可行的策略,以提高在低资源场景LLM性能.
- SynEL及其生成代码的公开发布有助于进一步研究和开发基于LLM的知识提取.
相关概念视频
Ligand Binding and Linkage
5.5K
Allosteric proteins have more than one ligand binding site; the binding of a ligand to any of these sites influences the binding of ligands to the other sites. When a protein is allosteric, its binding sites are called coupled or linked. In the case of enzymes, the site that binds to the substrate is known as the active site and the other site is known as the regulatory site. When a ligand binds to the regulatory site, this leads to conformational changes in the protein that can influence...
5.5K
Ligand Binding and Linkage
4.0K
4.0K
Bonding and Strength of Aggregate
458
The bond between aggregate particles and the cement matrix is significantly influenced by the shape and surface texture of the aggregates. High-strength concretes benefit from a rougher texture, which leads to stronger bonding due to greater adhesion. Angular aggregates with larger surface areas also enhance this bond. The bonding quality, however, is complex to assess as no universally accepted test exists. Good bonding is indicated when a crushed concrete specimen shows some aggregate...
458
Deleterious Substances in Aggregate
530
Deleterious substances in aggregates can be detrimental to the quality and durability of concrete. These substances include organic impurities like loam, which interfere with cement hydration and are usually present in the sand. These prevent a good bond between aggregate and cement paste. Organic impurities can be detected using the colorimetric test, where the darkness of a solution after agitation indicates the level of organic content.
Another type of impurity is clay and fine material that...
Another type of impurity is clay and fine material that...
530
Synthetic Biology
5.5K
Synthetic biology is an interdisciplinary science that involves using principles from disciplines such as engineering, molecular biology, cell biology, and systems biology. It involves remodeling existing organisms from nature or constructing completely new synthetic organisms for applications such as protein or enzyme production, bioremediation, value-added macromolecule production, and the addition of desirable traits to crops, to name a few.
Golden rice
Golden rice is a genetically modified...
Golden rice
Golden rice is a genetically modified...
5.5K
Phosphodiester Linkages
110.0K
Overview
Phosphodiester bond forms when a phosphoric acid molecule (H3PO4) links with two hydroxyl groups (–OH) of two other molecules, forming two ester bonds. Two water molecules are released in this process. The phosphodiester bond is commonly found in nucleic acids (DNA and RNA) and plays a critical role in their structure and function.
Phosphodiester Bonds Link Nucleotides Together
DNA and RNA are polynucleotides or long chains of nucleotides that are linked together. A nucleotide is...
Phosphodiester bond forms when a phosphoric acid molecule (H3PO4) links with two hydroxyl groups (–OH) of two other molecules, forming two ester bonds. Two water molecules are released in this process. The phosphodiester bond is commonly found in nucleic acids (DNA and RNA) and plays a critical role in their structure and function.
Phosphodiester Bonds Link Nucleotides Together
DNA and RNA are polynucleotides or long chains of nucleotides that are linked together. A nucleotide is...
110.0K

![Technical Aspect of the Automated Synthesis and Real-Time Kinetic Evaluation of [11C]SNAP-7941](/_next/image?url=https%3A%2F%2Fcloudfront.jove.com%2FCDNSource%2Fteasers%2F59557.jpg&w=3840&q=50)
