The Pitfalls of KV Cache Compression
This addresses practical deployment issues for practitioners using KV cache compression in LLMs, though it is incremental as it builds on existing compression methods.
The paper identifies pitfalls in KV cache compression for LLMs, showing that certain instructions degrade rapidly and can be ignored, with system prompt leakage as a case study, and proposes simple eviction policy changes to improve multi-instruction task performance.
KV cache compression promises increased throughput and efficiency with negligible loss in performance. While the gains in throughput are indisputable and recent literature has indeed shown minimal degradation on particular benchmarks, in general the consequences of compression in realistic scenarios such as multi-instruction prompting have been insufficiently studied. In this paper, we identify several pitfalls practitioners should be aware of when deploying KV cache compressed LLMs. Importantly, we show that certain instructions degrade much more rapidly with compression, effectively causing them to be completely ignored by the LLM. As a practical example of that, we highlight system prompt leakage as a case study, empirically showing the impact of compression on leakage and general instruction following. We show several factors that play a role in prompt leakage: compression method, instruction order, and KV eviction bias. We then propose simple changes to KV cache eviction policies that can reduce the impact of these factors and improve the overall performance in multi-instruction tasks.