Research · JUN 2026

Still: Amortized KV Cache Compaction in a Single Forward Pass

You can shrink a language model's KV cache by 200×, in a single forward pass, and it still answers correctly.

Abstract: The KV cache is the memory bottleneck of long-horizon language model deployment. Practically, a deployable compactor must be lightweight enough to call during inference, expressive enough to preserve context under constraint, and reusable across a trajectory. Existing compaction methods satisfy only part of this requirement: selection methods are lightweight but subset-bound, while synthesis methods are expressive but rely on per-context optimization. Here we introduce Still, a small per-layer Perceiver trained once against a frozen base model that produces compact keys and values in a single forward pass. On Qwen and Gemma models, Still occupies the favorable side of the speed--quality frontier across compression ratios from 8× to 200× and context lengths from 8k to 128k. On the long-context RULER grid, Still exceeds the strongest baseline by 8--22 points. The same compact cache also supports free-form summarization, preserving most of the full-context gain on HELMET and winning a pairwise LongBench summarization comparison against KV-Distill. Because compaction is a forward pass, Still can be applied iteratively, entering a long-horizon regime unavailable to per-context methods. We show that amortization makes long-context cache compaction tractable, and synthesis makes its compact state useful at extreme compression.

Read the paper

Cite this work

@article{iverson2026still,
  title   = {Still: Amortized KV Cache Compaction in a Single Forward Pass},
  author  = {O'Neill, C. and Sandomirsky, A. and Partridge, H. and Jayasekara, M. and Kirkby, M.},
  journal = {Base Labs},
  year    = {2026},
  number  = {001},
  url     = {https://baselabs-5x6p03ykz-blueprintbyb10.vercel.app/001},
}

Related from the lab