REVIEW 2 cited by
Q-Filters: Leveraging QK Geometry for Efficient KV Cache Compression
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
Autoregressive language models rely on a Key-Value (KV) Cache, which avoids re-computing past hidden states during generation, making it faster. As model sizes and context lengths grow, the KV Cache becomes a significant memory bottleneck, which calls for compression methods that limit its size during generation. In this paper, we discover surprising properties of Query (Q) and Key (K) vectors that allow us to efficiently approximate attention scores without computing the attention maps. We propose Q-Filters, a training-free KV Cache compression method that filters out less crucial Key-Value pairs based on a single context-agnostic projection. Contrarily to many alternatives, Q-Filters is compatible with FlashAttention, as it does not require direct access to attention weights. Experimental results in long-context settings demonstrate that Q-Filters is competitive with attention-based compression methods such as SnapKV in retrieval tasks while consistently outperforming efficient compression schemes such as Streaming-LLM in generation setups. Notably, Q-Filters achieves a 99% accuracy in the needle-in-a-haystack task with a x32 compression level while reducing the generation perplexity drop by up to 65% in text generation compared to Streaming-LLM.
Forward citations
Cited by 2 Pith papers
-
KVDiagnosis: A Diagnostic Benchmark for KV-Cache Compression in Long-Context Language Models
KVDiagnosis is a diagnostic benchmark that pairs 59,800 compressed runs with FullCache controls and attaches cache, likelihood, attention, and decoding measurements to 12,520 failure rows, finding 63.2% of failures ha...
-
RCStat: A Statistical Framework for using Relative Contextualization in Transformers
RCStat uses pre-softmax attention logits to define a Relative Contextualization score that improves adaptive KV-cache eviction and attention-head selection for attribution on LLaMA models.
Discussion (0). Continue with ORCID to comment.