OVAL constructs output-aware page bases for KV cache retrieval by minimizing a hybrid of key fidelity and value-coupled output sensitivity, improving over key-only spectral summaries.
Razorattention: Efficient KV cache compression through retrieval heads
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 2roles
background 1polarities
background 1representative citing papers
RTPurbo converts full-attention LLMs to sparse attention by retaining full KV for retrieval heads and using a low-dimensional dynamic indexer, achieving near-lossless accuracy after minimal adaptation.
citing papers explorer
-
OVAL: Output-Aware Local Page Bases for KV Cache Retrieval
OVAL constructs output-aware page bases for KV cache retrieval by minimizing a hybrid of key fidelity and value-coupled output sensitivity, improving over key-only spectral summaries.
-
Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps
RTPurbo converts full-attention LLMs to sparse attention by retaining full KV for retrieval heads and using a low-dimensional dynamic indexer, achieving near-lossless accuracy after minimal adaptation.