ParisKV: Fast and Drift-Robust KV-Cache Retrieval for Long-Context LLMs 事件

ParisKV: Fast and Drift-Robust KV-Cache Retrieval for Long-Context LLMs arXiv:2602.07721v3 Announce Type: replace-cross Abstract: KV-cache retrieval is essential for long-context LLM inference, yet existing methods struggle with distribution drift and high latency at scale. We introduce ParisKV, a drift-robust, GPU-native KV-cache retrieval framework based on collision-based candidate selection, followed by a quantized inner-product reranking estimator. For million-token contexts, ParisKV suppo