Jev 正式发布:面向决策的 System One 模型
TypeSafe 创始人 Diogo Almeida 介绍 Jev、RLCD 训练方法,以及面向软件决策场景的模型定位。原帖附有 2 分 56 秒发布视频。

Cris Lenta发布对比结果:lifesim-social-1在社交决策上准确率88%、延迟102ms、校准误差0.047,全面优于OpenAI Decisions API。
we beat OpenAI's Decisions API 24h after it launched it's Jev competitor lifesim-social-1 vs OpenAI Decisions on social decisions: • accuracy: 88% vs 67% • p50 latency: 102 ms vs 278 ms • calibration error: 0.047 vs 0.170 (lower is better) this means agents will not only