AIbase
EN
Home
Bleurt Large 128
B

Bleurt Large 128

Developed by Elron
BLEURT is a metric model for evaluating the quality of generated text, based on the pre-trained BERT architecture, capable of automatically assessing the similarity between generated text and reference text.
Text Embedding
Transformers
#Text Generation Evaluation#Multi-reference Scoring#Robust Metrics
Downloads 58
Release Time : 3/2/2022

Model Overview

BLEURT is a text generation evaluation metric model developed by Google Research, which evaluates generation quality by learning the similarity between reference texts and candidate texts. This PyTorch version is a converted implementation of the original BLEURT model.

Model Features

Robust Evaluation Metric
By learning from extensive human-rated data, it provides automatic evaluation results highly consistent with human judgment.
BERT-based Architecture
Utilizes the pre-trained BERT model to capture deep semantic features of text.
End-to-End Training
The entire model (including pre-training and fine-tuning phases) is optimized end-to-end for the evaluation task.

Model Capabilities

Automatic text quality evaluation
Similarity scoring between generated text and reference text
Multi-candidate text comparison

Use Cases

Natural Language Processing Research
Machine Translation Evaluation
Evaluating the quality of translations generated by different machine translation systems
Provides automatic scores highly correlated with human evaluation
Text Summarization Evaluation
Comparing the similarity between summaries generated by different summarization systems and reference summaries
Can replace some manual evaluation work
Model Development
Generative Model Tuning
Used as a loss function or evaluation metric for training generative models
Helps improve the output quality of generative models
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase