# Preprint reports workspace tools improve local AI data-engineering results

_Published Saturday, November 14, 2026 at 3:06 PM EST · AI, Science · Latest · Tier 2 — Notable_

A preprint reports that giving locally deployed language models a workspace with tools raised pass rates by 24.0 to 65.3 points over uninformed one-shot generation for models above two billion parameters. Researchers evaluated fifteen mobility data-engineering tasks in 3,600 scored attempts on a consumer-grade GPU.

A 9-billion-parameter model reached an 80.0% pass rate within about 6.5 GB, while the highest-scoring configuration reached 98.7%. Deterministic checkers assessed the scripts, tables, structured files, figures and reports the models produced. The benchmark covers mobility workflows and the evaluated configurations; its results do not establish performance across all data-engineering work.

## Sources

- [cs.AI updates on arXiv.org](https://arxiv.org/abs/2610.11482)

---
Canonical: https://techandbusiness.org/newswire/nfP-X-vaC2BYQnXwLrjSGM
Published: 2026-11-14T20:06:33.518Z
Story chronology: 2026-11-14T05:00:00.000Z
Retrieved: 2026-11-15T00:01:48.839Z
Publisher: Tech & Business (techandbusiness.org)