Loading…
HiChunk is a hierarchical document chunking framework for retrieval-augmented generation (RAG) systems that dynamically adjusts the semantic granularity of retrieval fragments. It ships with HiCBench, a benchmark to evaluate chunking quality using annotated documents to synthesize evidence-intensive QA pairs. The Auto-Merge retrieval algorithm complements HiChunk to mitigate information gaps arising from chunking, improving retrieval performance.
AI named HiChunk from January 2026 to March 2026.