Scholar Feed

KV-cache compression

PromptCache

Prompt Cache: Modular Attention Reuse for Low-Latency Inference

Superseded baseline#60 of 234 most-superseded · first seen Nov 7, 2023

Superseded — cited as a baseline and beaten by newer methods

2 papers critique it · 1 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites PromptCache as a baseline.

block-based methods like PromptCache~Gim2023PromptCM, TurboRAG~lu2024turborag, and BlockAttention~ma2024block ignore inter-block dependencies, causing performance degradation
— TableCache: Primary Foreign Key Guided KV Cache Precomputation for Low Latency Text-to-SQL
“PromptCache~gim2024prompt stores text segments that appear frequently on an inference server using a schema, although this approach hampers usability, as users must conform their natural language to the schema format.”
— SemShareKV: Efficient KVCache Sharing for Semantically Similar Prompts via Token-Level LSH Matching

Beaten on benchmarks

Head-to-head results where a newer method reports beating PromptCache. Values are copied from the source paper's tables — verify against the cited paper.

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.