haoailab· @haoailab · X·· 4 小时前AI 评分
GPT6-Astra 对抗有组织谎言测试
4️⃣ 但当证人被收买时… 💰 在 100 个更难的案例中,付费证人多于诚实证人。以下是各团队的表现: • 🏅 GPT6-Astra + Jev swarm:93 • 🥈 GPT6-Astra 单独:59 • 🥉 Jev swarm 单独:42 • ❓ 基于规则的领导者 + Jev swarm:0–1 要点:对抗有组织的谎言,你需要一个会追问"谁在说话"的推理领导者。
4️⃣ But when the witnesses are bribed… 💰
In 100 harder cases, paid witnesses outnumber honest ones. Here is how each team performs:
• 🏅 GPT6-Astra + Jev swarm: 93
• 🥈 GPT6-Astra alone: 59
• 🥉 Jev swarm alone: 42
• ❓ Rule-based leader + Jev swarm: 0–1
Takeaway: against organized lies, you need a reasoning leader who asks who's talking.
来源:haoailab · x.com