AIbase
EN
Home
Kobart
K

Kobart

Developed by hyunwoongko
KoBART-base-v2 is a Korean pre-trained model based on the BART architecture, optimized for long-sequence semantic processing through the addition of chat data.
Large Language Model
Transformers
 KoreanOpen Source License:MIT#Korean text processing#Long sequence semantic understanding#Chat data optimization
Downloads 72.70k
Release Time : 3/2/2022

Model Overview

This model is a Korean version of the BART architecture, specifically optimized for Korean text processing and capable of handling longer sequence semantics than the original KoBART.

Model Features

Long sequence processing optimization
Enhanced ability to handle long-sequence semantics through training with chat data.
Post-processor improvements
Added bos/eos post-processor and removed token_type_ids, improving processing efficiency.

Model Capabilities

Korean text generation
Korean text understanding
Long sequence semantic processing

Use Cases

Text processing
Sentiment analysis
Used for sentiment analysis tasks on Korean text
Achieved an accuracy of 0.901 on the NSMC dataset
Chat applications
Suitable for text generation in Korean chatbots
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase