Embedding similarity scores stuck around 0.541 when using BGE-Code-V1 as retriever in RAG

[OpEnD17]

When I use BGE-Code-V1 (Qwen2.5-Coder-1.5B based) as retriever in my RAG pipeline, I find that query–chunk similarity scores are always around ~0.541, regardless of the query and document content.

Task: RAG retrieval (query: incomplete code; chunk: code blocks(almost function-level)) Similarity function: cosine similarity