McGill-NLP / llm2vec

Code for 'LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders'
https://mcgill-nlp.github.io/llm2vec/
MIT License
1.17k stars 88 forks source link

Inquiry Regarding Instruction Addition to Query Statements #88

Closed shrijayan closed 4 months ago

shrijayan commented 4 months ago

I noticed that you have been adding instructions to query statements in your examples and code snippets. However, this detail wasn't mentioned in your paper. Could you please explain the rationale behind this practice? I found that including instructions with queries resulted in higher correlation scores between vectors compared to not using instructions. Understanding the reasoning and methodology would be very helpful.