QCFuse speeds up RAG inference in LLMs by 40% through query-focused cache fusion and selective token recomputation while preserving accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DB 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
QCFuse: Query-Centric Cache Fusion for Efficient RAG Inference
QCFuse speeds up RAG inference in LLMs by 40% through query-focused cache fusion and selective token recomputation while preserving accuracy.