# Preprint reports bounded-context gains for grid-model AI queries

_Published Thursday, September 3, 2026 at 12:07 PM EDT · Science · Latest · Tier 2 — Notable_

Authors of an arXiv preprint report a seed-anchored graph-rendering method for LLM question answering over CIM and CGMES power-grid models. On a preregistered 100-item SmallGrid bank under an 8,000-character context budget, reported accuracy rose from 0.450 to 0.970. The authors also report matching or exceeding certain extracted graph representations without LLM graph-construction tokens. The result is limited to the evaluated models, reader and context budget.

## Sources

- [cs.AI updates on arXiv.org](https://arxiv.org/abs/2609.02011)

---
Canonical: https://techandbusiness.org/newswire/v60iXtTE-h591cXK7fqt7M
Published: 2026-09-03T16:07:26.468Z
Story chronology: 2026-09-03T04:00:00.000Z
Retrieved: 2026-10-19T01:51:25.155Z
Publisher: Tech & Business (techandbusiness.org)