System One模型新趋势：CLM对比Jev在自定义Harness中的应用
Pay attention to this new wave of System One models if you are building custom h…
Pay attention to this new wave of System One models if you are building custom harnesses.
如果你正在构建自定义 harness，请密切关注这一波 System One 模型的新浪潮。
First Jev. Now, Contrastive Language Model (CLM).
首先是 Jev。现在，对比语言模型（CLM）。
CLM is 9x faster than Jev.
CLM 的速度比 Jev 快 9 倍。
CLM seems to be a better verifier than Jev, particularly at long-horizon tasks.
CLM 似乎比 Jev 是更好的验证器，特别是在长周期任务中。
How do Jev and CLM differ?
Jev 和 CLM 有何不同？
CLM is contrastive, and Jev is trained with Reinforcement Learning for Calibrated Decisions (RLCD).
CLM 是对比式的，而 Jev 是使用用于校准决策的强化学习（RLCD）训练的。
Jev receives a situation plus predefined questions, and returns typed decisions with probabilities. CLM embeds the situation and candidate actions, compares their similarity, then ranks or selects the best match.
Jev 接收情境加上预定义的问题，并返回带有概率的类型化决策。CLM 嵌入情境和候选动作，比较它们的相似度，然后对最佳匹配进行排序或选择。
The point is that there are several ways to attack this problem, which is exciting.
关键在于有多种方法可以解决这一问题，这令人兴奋。
You can see my recent guide on combining System One and System Two models for building custom harnesses. https://academy.dair.ai/resources/jev-decisions-in-a-pi-sdk-harness
你可以查看我最近关于结合 System One 和 System Two 模型以构建自定义 harness 的指南。https://academy.dair.ai/resources/jev-decisions-in-a-pi-sdk-harness
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力