Overview
Understanding Fine-Tuning of Large Language Models: A Comprehensive Overview
Fine-tuning Large Language Models (LLMs) is crucial for customizing AI to meet specific business needs. This blog delves into the two primary types of fine-tuning: instruction tuning, which enhances a model's ability to follow complex commands, and alignment tuning, which ensures outputs align with human values. By understanding these processes, businesses can effectively leverage AI for customer support, content creation, and more.
SAAS - Solving Ability Amplification Strategy for Enhanced Mathematical Reasoning in Large Language Models
Introducing "SAAS: Solving Ability Amplification Strategy for Enhanced Mathematical Reasoning in Large Language Models," a groundbreaking research paper by Upstage AI, Mathpresso Inc., and KT Corp. This novel approach leverages sequential learning, combining Chain-of-Thought (CoT) and Program-of-Thought (PoT) methodologies to significantly boost the mathematical reasoning and problem-solving skills of Large Language Models (LLMs). Our findings show that SAAS achieves state-of-the-art performance on benchmarks like GSM8K and MATH, outperforming larger models and setting new standards in AI-driven mathematical reasoning. Discover how SAAS can elevate your LLM’s capabilities.
Next Generation for AGI: Upstage’s On-Device LLM, WriteUp
Experience the power of AI at your fingertips with Upstage's on-device LLM, WriteUp. Enjoy AI assistance for writing tasks without the need for an internet connection, thanks to advanced model optimization and quantization technologies. Perfect for remote locations, WriteUp ensures data privacy and high performance on your personal device.
Introducing Solar Mini Chat ja: Expanding Language Support to Japanese
We are thrilled to announce that Solar mini chat now includes Japanese, alongside English and Korean. Specially fine-tuned for multi-turn chat, Solar mini chat ja excels in Japanese language interactions, offering high performance and an enhanced user experience. Ideal for applications demanding nuanced and context-aware communication, it surpasses many open-source models in key NLP tasks. Seamlessly integrate it with your existing API keys and elevate your Japanese chat applications!
Breaking Barriers: Revolutionize Your Work with Our Next-Level Embedding Model
Experience the Next Leap in Embedding Technology with Solar Embedding-1-Large: Our groundbreaking Solar Embedding-1-Large model is set to transform your work processes. With superior performance compared to OpenAI's models and a commitment to tackling even the toughest tasks, it's time to elevate your search systems and beyond.
Evalverse: Revolutionizing Large Language Model Evaluation with a Unified, User-Friendly Framework
Discover Evalverse: A groundbreaking framework revolutionizing Large Language Model evaluation. With its unified approach and user-friendly features, Evalverse simplifies assessment, making AI advancements inclusive and comprehensive. Explore its key features and architecture, and witness its practical application in our demonstrative video. Join us in driving innovation and accessibility in AI technology with Evalverse!
Open Source All About Data Processing, Dataverse
Dataverse is a freely-accessible open-source project designed to streamline the extract, transform, and load (ETL) pipeline using Python. In this post, we delve into the origins of this project and shed light on its future prospects in the realm of open-source data processing.
(Almost) Zero Hallucination with RAG and Groundedness Check
Unraveling the mechanics behind translating LLM outputs into leaderboard scores. Exploring a spectrum of evaluation mechanisms that span from automated metrics to human assessments rooted in real-world applicability.
LLM Evaluation Part2. Mechanics Behind LLM Scoring Systems
Unraveling the mechanics behind translating LLM outputs into leaderboard scores. Exploring a spectrum of evaluation mechanisms that span from automated metrics to human assessments rooted in real-world applicability.
LLM Evaluation Part1. What is a Benchmark Dataset?
Want to know why and how we evaluate LLM models?
![[2023 AI KOREA GRAND PRIZE] Upstage Wins AI Technology Award (Minister of Science and ICT Award)](https://images.squarespace-cdn.com/content/v1/659384103b38c97cdaf368bd/1721624115335-9F99C40W5V0UPSA9YLEA/AI%2BKOREA_WEB%2BHERO.jpg)
[2023 AI KOREA GRAND PRIZE] Upstage Wins AI Technology Award (Minister of Science and ICT Award)
Upstage wins ‘AI Technology Award,’ the highest award given by the Minister of Science and ICT, at the 2023 AI Korea Awards Event!
Reinterpreting the History of NLP-based AI through a Data-Centric Perspective
What insights can be gained by examining natural language processing (NLP) through a data-centric perspective? Explore our blog post that delves into the 70-year history of AI, covering rule-based systems, machine learning, deep learning, and the recent emergence of large language models.
Data-Centric AI in the Real World
Just like a car needs fuel to move and a recipe requires ingredients to make a meal, artificial intelligence systems also need their own kind of fuel and materials, which is data. Explore the practical applications of data in the real world through this blog.
Until the birth of OCR that recognizes text (Upstage in-house OCR image data collection challenge)
Discover how Upstage builds its high-performance OCR solution, "Document AI," through our in-house image data collection event. Gain insights into the importance of data for AI model training and understand why it's a crucial component for success.