Y. Tian, “KeepKV: Achieving Periodic Lossless KV Cache Compression for Efficient LLM Inference”, AAAI, vol. 40, no. 39, pp. 33259–33267, Mar. 2026.