RTN (KV-cache compression): superseded — cited as a baseline and beaten by newer methods. 0 paper(s) critique it, 4 beat it on benchmarks — #39 of 234 most-superseded. Sub-problem: cluster led by KIVI. Newer alternatives in the same sub-problem include SpectrumKV, Hurwitz Quaternion Multiplicative Quantization (HQMQ), OSCAR, OScaR, TriAxialKV.

KV-cache compression

RTN

Superseded baseline#39 of 234 most-superseded

Superseded — cited as a baseline and beaten by newer methods

0 papers critique it · 4 beat it on benchmarks

Beaten on benchmarks

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

DecoQuant beats RTN
35.9 vs 2.3
Average · [activations only, W 16-A 2]
Unlocking Data-free Low-bit Quantization with Matrix Decomposition for KV Cache Compression
MiniCache beats RTN
35.44 vs 4.90
Average · [Llama-2-7B-Chat]
MiniCache: KV Cache Compression in Depth Dimension for Large Language Models
SKVQ beats RTN
37.50 vs 6.76
Average · [Llama-2-7B-chat, Group-size 128, key-cache 2bit, value-cache 2bit, window-size 128]
SKVQ: Sliding-window Key and Value Cache Quantization for Large Language Models
WindowQuant beats RTN
48.5 vs 44.4
Avg · [LLaVA-v1.5-7B]
WindowQuant: Mixed-Precision KV Cache Quantization based on Window-Level Similarity for VLMs Inference Optimization

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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.