Tag
The paper introduces KVFetch, a training-free framework that adds a temporal recall channel to score-based KV cache compressors to fix "sequential forgetting" (verbatim copying breaking under compression), recovering copying performance from 0.8 to 78.4 and boosting the RULER-16K average by +8.4 at iso-budget.
TaskPress introduces a query-agnostic KV cache compression framework that uses a task guide as a meta-query and quantization scale factors to prune irrelevant tokens, enabling reusable caches across diverse queries with negligible overhead.
KVpop introduces a learned KV cache eviction policy supervised by future-attention targets, achieving high compression rates (e.g., 98% performance at 75% compression) on Qwen3 models while maintaining quality.