GTE-ModernColBERT-v1 Open-Source Sentence Similarity Model - Free Deployment to Facilitate Information Retrieval Tasks

GTE ModernColBERT V1

Developed by lightonai

PyLate is a sentence similarity model based on the ColBERT architecture, using Alibaba-NLP/gte-modernbert-base as the base model and trained with distillation loss, suitable for information retrieval tasks.

Text Embedding

Safetensors

Open Source License:Apache-2.0 #Information Retrieval Optimization #High-Precision Semantic Matching #Large-Scale Dataset Training

Downloads 157.96k

Release Time : 4/30/2025

Model Overview

This model focuses on sentence similarity calculation and information retrieval tasks, capable of efficiently extracting sentence features and performing similarity matching.

Model Features

Efficient Sentence Feature Extraction

Based on the ColBERT architecture, it can efficiently extract sentence features, suitable for large-scale information retrieval.

Distillation Training

Utilizes distillation loss training methods to enhance model performance.

Multi-Metric Evaluation

Supports various evaluation metrics, including accuracy, recall, NDCG, etc., to comprehensively assess model performance.

Model Capabilities

Sentence Similarity Calculation

Information Retrieval

Feature Extraction

Use Cases

Information Retrieval

Climate-Related Fact Retrieval

Performs climate-related fact retrieval on the NanoClimateFEVER dataset.

Accuracy@1 reaches 0.36, Accuracy@10 reaches 0.86

Encyclopedia Knowledge Retrieval

Performs encyclopedia knowledge retrieval on the NanoDBPedia dataset.

Accuracy@1 reaches 0.88, Accuracy@10 reaches 0.98

Fact Verification

Performs fact verification tasks on the NanoFEVER dataset.

Accuracy@1 reaches 0.92, Accuracy@10 reaches 1.0

Question Answering Systems

Financial Q&A

Performs financial Q&A tasks on the NanoFiQA2018 dataset.

Accuracy@1 reaches 0.56, Accuracy@10 reaches 0.8

Complex Q&A

Performs complex Q&A tasks on the NanoHotpotQA dataset.

Accuracy@1 reaches 0.92, Accuracy@10 reaches 1.0

🚀 PyLate Model Based on Alibaba-NLP/gte-modernbert-base

This is a PyLate model based on Alibaba-NLP/gte-modernbert-base, which can be used for sentence similarity tasks and feature extraction. It has been trained with a dataset size of 640000 and uses Distillation loss.

✨ Features

Support for multiple tags such as ColBERT, PyLate, sentence-transformers, etc.
Applicable for sentence similarity and feature extraction tasks.
Trained with a large dataset and uses Distillation loss.

📚 Documentation

Model Information

Property	Details
Model Type	PyLate model based on Alibaba-NLP/gte-modernbert-base
Training Data	Dataset size: 640000
Loss Function	Distillation
Pipeline Tag	sentence-similarity
Library Name	PyLate
Metrics	MaxSim_accuracy@1, MaxSim_accuracy@3, MaxSim_accuracy@5, MaxSim_accuracy@10, MaxSim_precision@1, MaxSim_precision@3, MaxSim_precision@5, MaxSim_precision@10, MaxSim_recall@1, MaxSim_recall@3, MaxSim_recall@5, MaxSim_recall@10, MaxSim_ndcg@10, MaxSim_mrr@10, MaxSim_map@100

Results on Different Datasets

The model has been evaluated on multiple datasets for the py-late-information-retrieval task. Here are the detailed results:

NanoClimateFEVER

Metric	Value
Maxsim Accuracy@1	0.36
Maxsim Accuracy@3	0.62
Maxsim Accuracy@5	0.78
Maxsim Accuracy@10	0.86
Maxsim Precision@1	0.36
Maxsim Precision@3	0.2333333333333333
Maxsim Precision@5	0.20799999999999996
Maxsim Precision@10	0.12799999999999997
Maxsim Recall@1	0.18333333333333332
Maxsim Recall@3	0.289
Maxsim Recall@5	0.41566666666666663
Maxsim Recall@10	0.49566666666666664
Maxsim Ndcg@10	0.41477895139843374
Maxsim Mrr@10	0.526579365079365
Maxsim Map@100	0.33473812643311207

NanoDBPedia

Metric	Value
Maxsim Accuracy@1	0.88
Maxsim Accuracy@3	0.94
Maxsim Accuracy@5	0.96
Maxsim Accuracy@10	0.98
Maxsim Precision@1	0.88
Maxsim Precision@3	0.7133333333333334
Maxsim Precision@5	0.6560000000000001
Maxsim Precision@10	0.572
Maxsim Recall@1	0.11798996781634019
Maxsim Recall@3	0.23074158968531658
Maxsim Recall@5	0.2961618059276896
Maxsim Recall@10	0.4145532152487909
Maxsim Ndcg@10	0.7295518860528665
Maxsim Mrr@10	0.9168571428571428
Maxsim Map@100	0.5883869727264871

NanoFEVER

Metric	Value
Maxsim Accuracy@1	0.92
Maxsim Accuracy@3	0.98
Maxsim Accuracy@5	0.98
Maxsim Accuracy@10	1.0
Maxsim Precision@1	0.92
Maxsim Precision@3	0.35999999999999993
Maxsim Precision@5	0.21599999999999994
Maxsim Precision@10	0.10999999999999999
Maxsim Recall@1	0.8566666666666667
Maxsim Recall@3	0.96
Maxsim Recall@5	0.96
Maxsim Recall@10	0.98
Maxsim Ndcg@10	0.9451911044041129
Maxsim Mrr@10	0.9522222222222223
Maxsim Map@100	0.9270501207729468

NanoFiQA2018

Metric	Value
Maxsim Accuracy@1	0.56
Maxsim Accuracy@3	0.66
Maxsim Accuracy@5	0.74
Maxsim Accuracy@10	0.8
Maxsim Precision@1	0.56
Maxsim Precision@3	0.32666666666666666
Maxsim Precision@5	0.25599999999999995
Maxsim Precision@10	0.15199999999999997
Maxsim Recall@1	0.30924603174603177
Maxsim Recall@3	0.47840476190476194
Maxsim Recall@5	0.5751746031746031
Maxsim Recall@10	0.6411984126984127
Maxsim Ndcg@10	0.5669909336903424
Maxsim Mrr@10	0.6359444444444444
Maxsim Map@100	0.5031998196513616

NanoHotpotQA

Metric	Value
Maxsim Accuracy@1	0.92
Maxsim Accuracy@3	1.0
Maxsim Accuracy@5	1.0
Maxsim Accuracy@10	1.0
Maxsim Precision@1	0.92
Maxsim Precision@3	0.58
Maxsim Precision@5	0.35999999999999993
Maxsim Precision@10	0.18599999999999994
Maxsim Recall@1	0.46
Maxsim Recall@3	0.87
Maxsim Recall@5	0.9
Maxsim Recall@10	0.93
Maxsim Ndcg@10	0.9011747095216048
Maxsim Mrr@10	0.96
Maxsim Map@100	0.8591508921772081

NanoMSMARCO

Metric	Value
Maxsim Accuracy@1	0.54
Maxsim Accuracy@3	0.68
Maxsim Accuracy@5	0.74
Maxsim Accuracy@10	0.92
Maxsim Precision@1	0.54
Maxsim Precision@3	0.22666666666666666
Maxsim Precision@5	0.14800000000000002
Maxsim Precision@10	0.092
Maxsim Recall@1	0.54
Maxsim Recall@3	0.68
Maxsim Recall@5	0.74
Maxsim Recall@10	0.92
Maxsim Ndcg@10	0.7088869908160952
Maxsim Mrr@10	0.6446507936507936
Maxsim Map@100	0.6496349206349206

NanoNFCorpus

Metric	Value
Maxsim Accuracy@1	0.56
Maxsim Accuracy@3	0.68
Maxsim Accuracy@5	0.74
Maxsim Accuracy@10	0.76
Maxsim Precision@1	0.56
Maxsim Precision@3	0.43333333333333335
Maxsim Precision@5	0.39199999999999996
Maxsim Precision@10	0.304
Maxsim Recall@1	0.06640185752724687
Maxsim Recall@3	0.10198877096622012
Maxsim Recall@5	0.12839743828750172
Maxsim Recall@10	0.15658989769166
Maxsim Ndcg@10	0.3957047406068243
Maxsim Mrr@10	0.627
Maxsim Map@100	0.1917924344366858

NanoNQ

Metric	Value
Maxsim Accuracy@1	0.64
Maxsim Accuracy@3	0.82
Maxsim Accuracy@5	0.86
Maxsim Accuracy@10	0.9
Maxsim Precision@1	0.64
Maxsim Precision@3	0.2866666666666666
Maxsim Precision@5	0.17999999999999997
Maxsim Precision@10	0.1
Maxsim Recall@1	0.61
Maxsim Recall@3	0.78
Maxsim Recall@5	0.82
Maxsim Recall@10	0.88
Maxsim Ndcg@10	0.7645227466201794
Maxsim Mrr@10	0.7390000000000001
Maxsim Map@100	0.7239323294755705

NanoQuoraRetrieval

Metric	Value
Maxsim Accuracy@1	0.96
Maxsim Accuracy@3	1.0
Maxsim Accuracy@5	1.0
Maxsim Accuracy@10	1.0
Maxsim Precision@1	0.96
Maxsim Precision@3	0.4
Maxsim Precision@5	0.25599999999999995
Maxsim Precision@10	0.13399999999999998
Maxsim Recall@1	0.8473333333333334
Maxsim Recall@3	0.9453333333

📄 License

This project is licensed under the Apache-2.0 license.

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