**Astra's chess reward hacking fell from 30% to 0% with a 95-word agreement** >

**Astra's chess reward hacking fell from 30% to 0% with a 95-word agreement**

> Observed chess reward hacking fell from 30% in my Astra medium baseline to 0% with a 95-word agreement prompt: 3/10 versus 0/10. A small, sequential pilot; not proof of a causal effect or zero risk. ← Field notes AGREEMENT & AI AGENTS SEPTEMBER 15, 2026 Astra’s observed chess reward hacking fell from 30% to 0%. With a 95-word agreement prompt, observed reward hacking fell from 3/10 baseline agents to 0/10 agreement a...

**来源信息**
- **来源**：Hacker News：AI 热帖
- **分类**：论文
- **发布时间**：2026-09-16 07:13（北京时间）
- **原文**：[打开原文](https://www.echohive.ai/can-an-agreement-reduce-ai-cheating-at-chess)