Benchmarking which AI lies best, using a roleplaying game
ownworldman · reddit · 2026-11-06
A Reddit user shares an entertainment-style benchmark video: a roleplaying game where LLMs (mainly ChatGPT vs Claude per the thumbnail) try to deceive and out-bluff each other, ranking which model is the best liar and which is most gullible — a game-theoretic alternative to standard evals for observing deception and reasoning behavior.
More from Fun
- Pro-AI poster asks anti-AI critics: is a machine doing human painting 'slop'? — charis_ai · 2026-11-06
- AI Agents Placed 500 Uber Eats Orders That Resulted in Just 4 Trips — fkasummer · 2026-11-06
- OpenAI devrel star Tibo's reach collapses to a fifth as even colleagues tire of the constant hyping — teortaxesTex · 2026-11-06
- Mistral agent lead Thibault Sautiaux 'demoted'? His own reply: "I live in the basement now" — CtrlAltDwayne · 2026-11-06
- Florida deploys robot bunnies to lure invasive pythons that wiped out 99% of raccoons — IsForAt · 2026-11-06
- My favorite album of the year is made by a robot — dreamwieber · 2026-11-06