KorNAT (Korean National Alignment Test) is a benchmark dataset developed to assess the alignment of Large Language Models (LLMs) with South Korean social values and common knowledge. It evaluates models across two primary dimensions: Social Value Alignment: Measures how well...

A knowledge graph is a structured representation of information where entities (nodes) are connected by relationships (edges) to form a graph. It organizes and links data in a way that allows machines to understand, retrieve, and infer new information efficiently,...

Keras is an open-source deep learning framework designed to simplify the creation and deployment of neural networks. It provides an intuitive, user-friendly interface for building and training models while offering compatibility with powerful backends like TensorFlow, Theano, and Microsoft Cognitive...

Instruction Tuning is the process of refining large language models (LLMs) so that they better follow user-provided instructions or prompts. Rather than simply predicting the next word based on statistical patterns, instruction-tuned models are trained on curated datasets of prompts...

Human alignment refers to the process of designing and training AI systems to act in ways that align with human intentions, values, and ethical principles. The goal is to ensure that AI systems perform tasks safely, reliably, and in accordance...

In the context of AI and machine learning, hallucination refers to the generation of outputs by an AI model that are incorrect, fabricated, or nonsensical, despite appearing plausible or confident. This phenomenon is common in generative models, especially large language...

Ground truth refers to the accurate, verified data used as a benchmark or standard to evaluate and train machine learning models. It represents the true state or label of data, often collected through expert annotation, observation, or measurement, and is...

Graph Retrieval-Augmented Generation (Graph RAG) is an advanced methodology in AI that combines retrieval-augmented generation systems with graph-based data structures, such as knowledge graphs or graph databases. This approach enhances large language models (LLMs) by incorporating structured, relational data to...

GPT (Generative Pre-trained Transformer) is a type of large language model developed by OpenAI that uses a transformer-based architecture to generate human-like text. It is trained on vast amounts of text data in an unsupervised manner, learning patterns, context, and...

A golden dataset is a high-quality, meticulously curated dataset used as a benchmark or standard for evaluating, validating, and training machine learning models. It is often considered the “gold standard” for a specific task, containing accurate, consistent, and representative data...