HeatKV doubles KV-cache compression ratios over prior methods for VAR models by creating static head-specific pruning schedules from attention rankings on a calibration set, while preserving image quality on Infinity-2B.
Dispenser: Hierarchical KV Cache Management for Efficient LLM Generative Inference
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
No citing papers match the current filters.