gte-Qwen2-1.5B-instruct
| Property | Value |
|---|---|
| Parameter Count | 1.5B |
| Embedding Dimension | 1536 |
| Max Input Length | 32k tokens |
| License | Apache 2.0 |
| Paper | Research Paper |
What is gte-Qwen2-1.5B-instruct?
gte-Qwen2-1.5B-instruct is the latest addition to the General Text Embedding (GTE) model family, built on the Qwen2-1.5B architecture. This model represents a significant advancement in multilingual text embedding, achieving impressive performance across English, Chinese, French, and Polish languages while maintaining a relatively compact size compared to its larger 7B parameter counterpart.
Implementation Details
The model incorporates several sophisticated architectural elements including bidirectional attention mechanisms and instruction tuning specifically optimized for the query side. It's trained on a comprehensive multilingual corpus using both weakly supervised and supervised approaches, resulting in robust cross-lingual capabilities.
- Advanced bidirectional attention architecture
- Query-specific instruction tuning
- Comprehensive multilingual training
- Efficient 1536-dimensional embeddings
- Support for long sequences up to 32k tokens
Core Capabilities
- Strong performance on MTEB benchmark (67.16% average score)
- Excellent multilingual capabilities across C-MTEB (67.65%), MTEB-fr (66.60%), and MTEB-pl (64.04%)
- Efficient text embedding generation for retrieval and classification tasks
- Robust performance in semantic similarity tasks
Frequently Asked Questions
Q: What makes this model unique?
The model stands out for its excellent balance between model size and performance, offering near state-of-the-art results while being significantly smaller than 7B parameter models. It's particularly notable for maintaining strong performance across multiple languages and tasks.
Q: What are the recommended use cases?
The model excels in various applications including semantic search, document retrieval, text classification, and cross-lingual information retrieval. It's particularly well-suited for production environments where computational efficiency is important but high-quality multilingual embeddings are required.
