M

Meltemi 7B Instruct V1.5

Developed by ilsp
Meltemi 7B Instruct v1.5 is a Greek instruction fine-tuned large language model improved based on Mistral 7B, focusing on Greek natural language processing tasks.
Downloads 1,237
Release Time : 7/31/2024

Model Overview

This model is an instruction fine-tuned large language model optimized for the Greek language, with the ability to efficiently process Greek text, supporting long context understanding and complex instruction execution.

Model Features

Greek vocabulary expansion
The Greek vocabulary of the Mistral 7b tokenizer has been expanded, significantly improving the Greek tokenization efficiency (from 6.80 tokens per word to 1.52 tokens).
Long context processing
Supports a context length of 8192 and can handle more complex text inputs.
ORPO fine-tuning algorithm
The Odds Ratio Preference Optimization algorithm is used for fine-tuning, and 97k preference data is used.
Performance improvement
Compared with the base model, the performance on the Greek test set has an average improvement of 7.8%.

Model Capabilities

Greek text generation
Greek question-answering system
Greek instruction understanding
Long text processing

Use Cases

Education
Greek learning assistance
Help students understand and learn Greek
Provide accurate Greek explanations and examples
Medical
Medical question-answering system
Answer Greek medical-related questions
Achieve an accuracy of 48% in the medical multiple-choice question-answering test
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase