Claude Mythos Preview Early Snapshot Achieves Over 2x Time Horizon vs. Next Best Model in METR Evaluation
Alex Albert(@alexalbert__)126 字 (约 1 分钟)
75
Claude Mythos Preview achieved a time horizon more than 2x that of the next best model on METR's 80% success rate benchmark, with a minimum 50%-time-horizon of 16 hours (95% CI: 8.5–55 hours).
入选理由：Claude Mythos Preview 时间跨度达 16 小时（95% CI 8.5–55 小时）
FeaturedTweet#Claude#Mythos#AI Evaluation#Time Horizon英文