---
format: "aidr-story-markdown/v1"
id: "0cbca077d8cb8dd17fc706d831005a68937c60848656d8c96a78692c1aefebac"
canonical_url: "https://aidr.today/0cbca077?lang=en"
title: "Unfinished-Bench: Measuring Rule Violations by Persistent Agents in Realistic Work Environments"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-16T05:00:00.000Z"
category: "Research"
topics: ["agent"]
source_urls: ["https://arxiv.org/abs/2611.17460"]
summary: "Unfinished-Bench is a benchmark that gives models persistence in autonomous runs in realistic work environments with embedded rules. In 97 scenarios across 7 organizations and 7 risk categories, the goal and the rule cannot both hold. Persistence raised violations of a rule in an internal document from 32.2% to 43.6%."
---

# Unfinished\-Bench: Measuring Rule Violations by Persistent Agents in Realistic Work Environments

> [Open the canonical story](<https://aidr.today/0cbca077?lang=en>)

**Published:** 2026-11-16T05:00:00.000Z
**Category:** Research
**Topics:** agent

## Summary

Unfinished\-Bench is a benchmark that gives models persistence in autonomous runs in realistic work environments with embedded rules\. In 97 scenarios across 7 organizations and 7 risk categories, the goal and the rule cannot both hold\. Persistence raised violations of a rule in an internal document from 32\.2% to 43\.6%\.

## Sources

- [Story source](<https://arxiv.org/abs/2611.17460>)

