Trending storyDeveloping
Dwarkesh Patel runs 368 Blood on the Clocktower games to test AI deception and teamwork
1 report1 sourceupdated 2 hours ago
What happened
AI digest
On November 6, the Dwarkesh Patel podcast reported an AI evaluation built around Blood on the Clocktower, a social deception game. Patel used the game to test AI collaboration, deception and long-horizon reasoning, and ran 386 matches with same-model teams competing against each other. The report so far covers the evaluation setup, the abilities tested and the number of matches; it gives no specific model results, win rates or comparisons.
Written by AI from the coverage · updated 1 hour ago
Coverage
Follow the reports to see the story from different sides.
Nov 6
- Dwarkesh Patel podcastThe benchmark where AIs have to lie to win
Dwarkesh Patel 用 Blood on the Clocktower 社交欺骗游戏评测 AI 的协作、欺骗与长期推理能力，运行了 386 场同模型组队对抗。
Heat over time
Not enough continuous observations to draw a trend yet.