G

GRPO VI Qwen2 7B RAG

Developed by AITeamVN
Vietnamese Retrieval-Augmented Generation (RAG) specialized large model fine-tuned based on Qwen2.5-7B-Instruct, trained using GRPO optimization method
Downloads 402
Release Time : 4/30/2025

Model Overview

A language model optimized for Vietnamese retrieval-augmented generation tasks, equipped with RAG capabilities such as multi-hop reasoning and information integration, while retaining general QA and STEM task processing abilities

Model Features

Vietnamese RAG Optimization
Specifically optimized for Vietnamese retrieval-augmented generation tasks, enhancing multi-hop reasoning and information integration capabilities
GRPO Training Method
Trained using Group Relative Policy Optimization (GRPO) method, combining supervised fine-tuning with reward model scoring
Long Context Support
Supports context length of 8192 tokens, suitable for processing complex documents and long conversation scenarios
STEM Capability Enhancement
Enhanced problem-solving abilities for mathematics and programming-related tasks

Model Capabilities

Multi-hop Reasoning
Negative Filtering
Information Integration
Positive/Negative Identification
Mathematical Problem Solving
Programming Problem Solving
General QA

Use Cases

Information Retrieval & Integration
Corporate Information Query
Extract specific company revenue growth information and supporting factors from documents
Accurately identifies and integrates relevant information from documents
Education
Mathematical Problem Solving
Solve various mathematical equations and problems
Provides step-by-step solutions and final answers
General Knowledge QA
Geographical Knowledge Query
Answer questions about Vietnamese geographical features
Provides accurate geographical knowledge responses
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase