Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
DiTyΒ 
posted an update 3 days ago
Post
92
Your documents have headings. Does your RAG chunker use them? πŸ‘€

I put together OutlineLM πŸ€—β€” a couple of models to help prepare meaningful chunks for RAG from Russian and English documents.

The focus is on headings that introduce a new semantic block, providing useful boundaries for splitting your documentsπŸ“„πŸ“„πŸ“„

Use the detected headings as chunk boundaries, apply size limits where needed, and pass the chunks to your retrieval pipeline.

πŸ”— Models & examples:
β€’ DiTy/OutlineLM-Gemma4-E2B-v1
β€’ DiTy/OutlineLM-Gemma4-E4B-v1

Give them a try β€” I’d love to hear how they handle your documents
In this post