Tag
CacheBlend, a EuroSys 2025 Best Paper, solves the problem that 90% of KV cache is never reused due to rigid prefix-matching in prompt caching. By selectively recomputing only boundary tokens between documents, it achieves 2-4x faster multi-document processing without quality loss, implemented in the open-source LMCache layer.
Parsewise is an API for agentic multi-document processing, enabling efficient handling of multiple documents.
CHOP is a framework for improving RAG systems on multi-document retrieval by using context-aware metadata and LLM-based chunk relevance evaluation to reduce semantic conflicts and hallucinations. The approach achieves 90.77% Top-1 Hit Rate through intelligent chunking and contextual preservation strategies.