← back to paper

Review history

arxiv: 2607.06519 · 2 revisions

FreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM Inference

1. 2026-07-11 CONDITIONAL MODERATE v1.1.0-grok45 novelty 5.0
   42224 ms 16852 in 3592 out 2026-07-11T00:10:34.383023+00:00
2. 2026-07-08 CONDITIONAL MODERATE v1.1.0-glm novelty 6.0
   567120 ms 15036 in 2571 out 2026-07-08T02:56:59.861933+00:00