H

Higgs Llama 3 70B

Developed by bosonai
A model post-trained based on Meta-Llama-3-70B, specifically optimized for role-playing tasks while excelling in general domain instruction following and reasoning tasks.
Downloads 220
Release Time : 6/5/2024

Model Overview

Higgs-Llama-3-70B is a post-trained model based on Meta-Llama-3-70B, specifically optimized for role-playing tasks while excelling in general domain instruction following and reasoning tasks. Through supervised fine-tuning and preference optimization, the model achieves alignment, enabling it to more closely adhere to character settings.

Model Features

Role-playing optimization
Specifically optimized for role-playing tasks, enabling closer adherence to character settings.
Instruction following
Excels in general domain instruction following tasks.
Reasoning capability
Performs excellently in reasoning tasks, surpassing similar models.
Alignment strategy
Employs special strategies to ensure model behavior aligns with system messages.

Model Capabilities

Text generation
Role-playing
Instruction following
Reasoning tasks

Use Cases

Role-playing
Imitating specific character dialogue
Imitating dialogue in the style of Sheldon Cooper
Can closely adhere to character settings, generating dialogue that matches the character's personality.
General instruction following
Answering complex questions
Answering complex questions from benchmarks like MMLU-Pro
Scores 63.2 on the MMLU-Pro test, outperforming Llama-3-70B-Instruct.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase