AIbase
EN
Home
Misted 7B
M

Misted 7B

Developed by Walmart-the-bag
Misted-7B is a 7B-parameter large language model based on the fusion of OpenHermes-2-Mistral-7B and Mistral-7B-SlimOrca, primarily designed for text generation tasks.
Large Language Model
Transformers
 EnglishOpen Source License:Apache-2.0#Multitask Text Generation#Few-shot Learning#Knowledge Reasoning
Downloads 386
Release Time : 11/13/2023

Model Overview

Misted-7B is a fusion model that combines the strengths of OpenHermes-2-Mistral-7B and Mistral-7B-SlimOrca, suitable for various text generation tasks including reasoning, Q&A, and dialogue generation.

Model Features

Model Fusion
Combines OpenHermes-2-Mistral-7B and Mistral-7B-SlimOrca using the slerp method, integrating their strengths.
Multitask Support
Supports various text generation tasks, including reasoning, Q&A, and dialogue generation.
Efficient Performance
Performs well on multiple benchmarks, especially on the HellaSwag and Winogrande datasets.

Model Capabilities

Text Generation
Reasoning Tasks
Q&A Systems
Dialogue Generation

Use Cases

Education
Reasoning Problem Solving
Used to solve reasoning challenges such as the AI2 Reasoning Challenge.
Achieved 63.65% accuracy on the AI2 Reasoning Challenge.
General Q&A
Common-sense Q&A
Answers common-sense-based questions.
Achieved 84.14% accuracy on the HellaSwag dataset.
Math Problems
Math Problem Solving
Solves math problems from the GSM8k dataset.
Achieved 59.59% accuracy on the GSM8k dataset.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
EnglishįŽ€äŊ“中文įšéĢ”中文ãĢãģんご
Š 2025AIbase