[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6},"actionData","errors","routes/research._index",{"_7":8},"items",[9],{"_10":11,"_12":13,"_14":15,"_16":17,"_18":19,"_20":21,"_22":23,"_24":25,"_26":27,"_28":29,"_30":31,"_32":33,"_34":33,"_35":33},"id","seed-research-context-compaction","type","research","slug","context-compaction","title","Context compaction for long-horizon agents","body_markdown","## Summary\n\nWe study how to keep long-running agent sessions within a fixed token budget without losing task-relevant context. The full writeup is in the [PDF](https://cdn.simse.dev/labs/context-compaction.pdf).\n\n## Results\n\nCompaction holds task accuracy within 2% of the uncompacted baseline while cutting context tokens by 60% on sessions longer than 50 turns.","body_html","

Summary

We study how to keep long-running agent sessions within a fixed token budget without losing task-relevant context. The full writeup is in the PDF.

Results

Compaction holds task accuracy within 2% of the uncompacted baseline while cutting context tokens by 60% on sessions longer than 50 turns.

\n","excerpt","A method for compacting conversation history that preserves task-relevant context while bounding token growth across long sessions.","status","published","tags",[48,49,50],"authors",[46,47],"meta",{"_36":23,"_37":38,"_39":40},"published_at","2026-05-20 00:00:00","created_at","updated_at","abstract","pdf_url","https://cdn.simse.dev/labs/context-compaction.pdf","links",{"_41":38,"_42":43,"_44":45},"paper","repo","https://github.com/simsedev/research-context-compaction","post","/blog/posts/hello-simse","A. Researcher","B. Engineer","context","memory","agents"]