← 研踪 PaperTrailarXiv:2611.20524 · v1
GaugeRoPE: RoPE-Compatible Executable Latent Attention
首次提交 2026-11-16 12:55
版本更新 2026-11-16 12:55
雷达观测 2026-11-17 13:07
摘要
Latent KV cache compression reduces memory footprint by mapping historical keys and values into compact representations. However, generic compressed key states require reconstruction to native dimension before applying RoPE and computing attention scores, adding sequence-scaling overhead to the decode path. We characterize when a latent KV representation can serve directly as the execution state of pretrained RoPE. For RoPE-based transformers, exact static linear execution of the rotary key path requires the latent-to-native map to preserve the pretrained RoPE action through an intertwining relation, restricting exact mixing to coordinates that share the same RoPE frequency. Guided by this characterization, we build GaugeRoPE, a training-free executable latent KV representation. It combines a RoPE-compatible carrier that preserves the rotary action with an approximate position-independent residual key--value (RKV) correction for complementary key and value information, both evaluated without per-token dense historical K/V reconstruction. On Mistral-7B, GaugeRoPE achieves KV compression while maintaining quality close to the full-cache baseline. Additional MHA results validate the construction beyond GQA. GaugeRoPE achieves up to the decode throughput of the vanilla uncompressed baseline through direct latent execution.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-16 12:55 · 最新观测版本
GaugeRoPE: RoPE-Compatible Executable Latent Attention
Latent KV cache compression reduces memory footprint by mapping historical keys and values into compact representations. However, generic compressed key states require reconstruction to native dimension before applying RoPE and computing attention scores, adding sequence-scaling overhead to the decode path. We characterize when a latent KV representation can serve directly as the execution state of pretrained RoPE. For RoPE-based transformers, exact static linear execution of the rotary key path requires the latent-to-native map to preserve the pretrained RoPE action through an intertwining relation, restricting exact mixing to coordinates that share the same RoPE frequency. Guided by this characterization, we build GaugeRoPE, a training-free executable latent KV representation. It combines a RoPE-compatible carrier that preserves the rotary action with an approximate position-independent residual key--value (RKV) correction for complementary key and value information, both evaluated without per-token dense historical K/V reconstruction. On Mistral-7B, GaugeRoPE achieves KV compression while maintaining quality close to the full-cache baseline. Additional MHA results validate the construction beyond GQA. GaugeRoPE achieves up to the decode throughput of the vanilla uncompressed baseline through direct latent execution.
查看此版本解读 →查看此版本原文 ↗