INSTRUCTOR

Overview

INSTRUCTOR is an instruction-finetuned text embedding model that generates domain-specific and task-aware embeddings without any fine-tuning, simply by providing task instructions.

Key Innovation

Unlike traditional embedding models with fixed representations, INSTRUCTOR computes embeddings based on instructions explaining the use case (e.g., "Represent the Science question for retrieving supporting documents").

Model Variants

• hkunlp/instructor-base: Base model
• hkunlp/instructor-large: Larger variant
• hkunlp/instructor-xl: Largest model with highest quality

Performance

State-of-the-art on 70 diverse embedding tasks (MTEB leaderboard)

Use Cases

• Classification with domain-specific embeddings
• Retrieval optimized for specific document types
• Clustering tailored to domain
• Text evaluation with custom criteria
• Cross-domain transfer without retraining

Availability

Hugging Face, LangChain, Haystack support