FastV (KV-cache compression): superseded — cited as a baseline and beaten by newer methods. 2 paper(s) critique it, 2 beat it on benchmarks — #42 of 234 most-superseded. Sub-problem: cluster led by ReKV. Newer alternatives in the same sub-problem include MuKV, KVCapsule, Decoupled Streaming Cache (DSCache), LightKV, Hierarchical Adaptive Eviction (HAE).

Superseded baseline#42 of 234 most-superseded

FastV

KV-cache compression

superseded — cited as a baseline and beaten by newer methods

2 papers critique it · 2 beat it on benchmarks

What papers say

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

“However, most existing methods depend on query-derived text tokens to compute attention scores and initiate compression, inevitably causing response delays in online scenarios.”
— LiveVLM: Efficient Online Video Understanding via Streaming-Oriented KV Cache and Retrieval
“FastV, though training-free, prunes vision tokens without cross-modality guidance, yielding inconsistent results across models and benchmarks.”
— Make Your LVLM KV Cache More Lightweight

Beaten on benchmarks

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

KVCapsule beats FastV · MME [Qwen2]
1634.22 vs 1498.46
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · LLaVABench [Qwen2]
74.80 vs 55.50
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · HallusionBench [Qwen2]
60.99 vs 58.68
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · LLaVABench [Qwen2.5]
83.70 vs 12.90
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · HallusionBench [Qwen2.5]
62.20 vs 25.60
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · MME [Qwen3]
1708.49 vs 561.25
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · MMMU [Qwen3]
0.48 vs 0.35
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · COCO [Qwen3]
15.80 vs 8.01
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · LLaVABench [Qwen3]
73.80 vs 22.30
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · MME [LLaVA-Mistral]
1458.10 vs 1440.69
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · MME [LLaVA-Llama]
1533.40 vs 1525.65
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy
KVCapsule beats FastV · MMMU [LLaVA-Llama]
0.60 vs 0.46
KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy

Newer alternatives

Recent methods in the same sub-problem, not yet superseded in the knowledge base.