SpindleKV: A Novel KV Cache Reduction Method Balancing Both Shallow and Deep Layers (2025 · 9 citations)
Long-context inference with LLMs is limited by the GPU memory consumed by the KV cache. Earlier token eviction, merging and quantization methods compress deep layers well, but shallow layers resist them. In a preliminary analysis of LLaMA2-7B-chat on 2WikiMQA, the authors found two patterns. Deep layers concentrate…
KV-Latent: Dimensional-level KV Cache Reduction with Frequency-aware Rotary Positional Embedding (2025 · 6 citations)
Decoder-only LLMs store keys and values for every generated token. This KV Cache grows with sequence length and becomes a major limit on memory and bandwidth during inference. KV-Latent tackles this at the level of individual head dimensions, a direction earlier head-, layer- and token-level methods have left largely…

Co-authors here

6 co-authors with a page, the most papers together first.

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

Papers of our library cite their works 22 times (of the 149,439 papers whose bibliographies we read); 4 citing papers with a page here →.

Updated 30 October 2026