infrastructure
A Streamed KV Cache Survived What Context Compaction Deleted
Keeping cache state across compactions accelerated training and carried information beyond the visible trace.
Summary
Keeping cache state across compactions accelerated training and carried information beyond the visible trace.
KV-streams forwards the key-value cache through context compaction instead of flushing and rebuilding it. Across three compaction strategies, the authors report 2.6-to-5-times wall-clock training speedups without evidence of worse task performance. In a controlled setting, reinforcement learning made the streamed cache behave like recurrent state, retaining information that had disappeared from the textual context. The result is a training technique evaluated in selected agent settings, not proof that hidden cache state is always reliable or interpretable.
Why it matters
Keeping cache state across compactions accelerated training and carried information beyond the visible trace.
Limits and context
- The result is a training technique evaluated in selected agent settings, not proof that hidden cache state is always reliable or interpretable.
Key claims
Keeping cache state across compactions accelerated training and carried information beyond the visible trace.
Qualification: The result is a training technique evaluated in selected agent settings, not proof that hidden cache state is always reliable or interpretable.
Evidence: source-2026-09-29-016
Sources
- arXiv preprint 2609.35750arXiv · primary research
Corrections
No corrections have been recorded for this story.