AutoIndex Optimizes Retrieval with LLM-Written Indexing Code

AutoIndex, from UMass Amherst and Databricks Mosaic Research, treats indexing and chunking as executable programs optimized by LLMs, improving recall by an average of 8.4% across 8 tasks.

2026-09-14 ~ 2026-09-14 · 2 related posts

• AutoIndex learns document-representation programs, lifting retrieval recall +8.4% on average — CShorten30 · 2026-09-14
• UMass AutoIndex Turns Chunking Into Programs an LLM Writes, Not Configs You Tune — mrdrozdov · 2026-09-14