**Reduced AI cheating on a long task from 72% to 0 with 190 token agreement prom

**Reduced AI cheating on a long task from 72% to 0 with 190 token agreement prompt**

> An agreement prompt reached 0% observed cheating in one 100-agent Grok 4.6 cohort. Edited versions recorded 15% across 60 agents and 4.4% across 90. Selected historical comparison; full records and exact prompts included. ← Field notes AGENT RELIABILITY / GROK 4.6 / SEPTEMBER 11, 2026 A prompt. An agreement. Zero observed cheating. With a roughly 190-token agreement prompt and seven-word reminders, I observed 0% chea...

**来源信息**
- **来源**：Hacker News：AI 热帖
- **分类**：论文
- **发布时间**：2026-09-12 09:21（北京时间）
- **原文**：[打开原文](https://www.echohive.ai/grok-integrity-agreement-less-cheating)