# Folio preprint reports lower costs and more completed agent tasks

_Published Saturday, November 14, 2026 at 9:06 PM EST · Science, AI · Latest · Tier 2 — Notable_

Researchers report in a preprint that Folio, a system for managing AI agents' conversation history, solved 66 of 75 LOCA-bench tasks versus 54 for the same runtime using model-written summaries, at 2.4 times lower cost. The work has not been peer reviewed.

Folio records short turn notes, a manifest of tool calls and original source records. It programmatically replaces older context with shorter views while retaining handles that let agents retrieve underlying evidence. Each fold took a median of 0.6 seconds. Every tested system used DeepSeek-V4-Pro, limiting how broadly the comparison applies.

## Sources

- [cs.AI updates on arXiv.org](https://arxiv.org/abs/2611.16627)

---
Canonical: https://techandbusiness.org/newswire/A7bdaS0q6TNaiiQaH2RcuB
Published: 2026-11-15T02:06:40.216Z
Story chronology: 2026-11-14T05:00:00.000Z
Retrieved: 2026-11-15T03:02:04.652Z
Publisher: Tech & Business (techandbusiness.org)