VikhrT5-3b Open-source Russian Model - Optimized Based on FLAN T5 3b, Outperforms FRED T5XL in Performance

Home

Vikhrt5 3b

Developed by Vikhrmodels

Russian-optimized model based on FLAN T5 3b, outperforming FRED T5XL

Large Language Model

Transformers

OtherOpen Source License:Apache-2.0 #Russian NLP optimization #Enhanced knowledge reasoning #Multi-task fine-tuning

Downloads 35

Release Time : 12/17/2023

Model Overview

VikhrT5-3b is a Russian large language model optimized based on the FLAN T5 3b architecture, specializing in Russian natural language processing tasks and demonstrating excellent performance across multiple Russian benchmarks.

Model Features

Russian optimization

Specially optimized for Russian, excelling in Russian language tasks

Performance advantage

Outperforms FRED-T5-XL and FLAN-t5-xl models in multiple Russian benchmarks

Large parameter scale

3b parameters provide stronger language understanding and generation capabilities

Model Capabilities

Russian text understanding

Russian text generation

Russian question answering

Russian reasoning

Use Cases

Natural language processing

Russian knowledge QA

Answering Russian knowledge-based questions

Accuracy of 0.32 on ru_mmlu dataset

Russian language understanding

Russian natural language understanding tasks

Accuracy of 0.4280 on xnli_ru dataset

Russian reasoning

Russian contextual reasoning tasks

Accuracy of 0.71 on xwinograd_ru dataset

Dataset	VikhrT5-3b	FRED-T5-XL(1.7b)	FLAN-t5-xl(3b)
ru_mmlu	0.32	0.252	0.28 (lol2)
xwinograd_ru	0.71 (lol)	0.57	0.52
xnli_ru	0.4280	0.34	0.33

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご

Vikhrt5 3b

Model Overview

Model Features

Model Capabilities

Use Cases

🚀 VikhrT5-3b: Enhanced Model Based on FLAN T5 3b for Russian

📄 License

📚 Documentation

Comparison Table

Datasets