Scholar Feed

KV-cache compression

LURE

Superseded baseline#180 of 234 most-superseded

Cited as a baseline — critiqued by newer work, not yet beaten on a benchmark here

1 papers critique it · 0 beat it on benchmarks

What papers say

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

some approaches involve additional training or use external knowledge to guide the model. LURE~zhou2023analyzing trains an extra state detector that triggers the regeneration of detected hallucinated contents by a revisor model
— PruneHal: Reducing Hallucinations in Multi-modal Large Language Models through Adaptive KV Cache Pruning

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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.