JEV 论文：廉价评审级联可省约 43% 评测成本
一篇讲怎么省评测钱的论文：便宜的 JEV 先判断，拿不准的才丢给 GPT-6，成本砍掉四成多，准确率几乎不掉。
TypeSafe AI 发布 Jev-as-a-Judge 论文，介绍只输出判定结果和标签概率、不含推理文本的决策型评审模型 JEV。在 510 个保留偏好对上，把 JEV 的置信判定留下、不确定案例升级给 GPT-6 Astra 的级联方案，保留了 GPT-6 约 99% 的准确率，费用约为其 57%。JEV 单价为每千次判定 0.044 美元、中位延迟 0.152 秒，比 GPT-6 的 12.182 美元和 1.885 秒便宜约 277 倍。在 RewardBench 上 JEV 拿到 92.2%（GPT-6 为 93.5%），HaluEval 上为 87.5%（GPT-6 为 86.7%），但 JudgeBench 上差距扩大到 78.6% 对 93.1%。作者提醒升级阈值并非对所有备用模型通用，建议在自己的数据上重新设定。
Banger paper introducing Jev-as-a-Judge.
The overall finding is that you want to use a cheap judge for most of your evals and send only the uncertain calls to a frontier model.
This paper measures how well that works with JEV, TypeSafe AI's decision-only judge.
On 510 held-out preference pairs, a cascade that accepted JEV's confident verdicts and escalated the rest to GPT-6 Astra kept 99% of GPT-6's accuracy at about 57% of its fee.
JEV returns a verdict and label probabilities with no reasoning text.
It costs $0.044 per 1,000 judgments at a median latency of 0.152 seconds, against $12.182 and 1.885 seconds for GPT-6, about 277 times cheaper.
On ordinary preference and evidence-grounded factuality it stays within 3 points of GPT-6 (92.2% against 93.5% on RewardBench, 87.5% against 86.7% on HaluEval).
The gap grows to 9 to 20 points on tasks that require checking a derivation or rejecting an elaborately written wrong answer, such as JudgeBench (78.6% against 93.1%).
On several benchmarks, JEV's gap to GPT-6 is concentrated in its low-confidence decisions, which is why the cascade works.
The escalation threshold did not transfer for every fallback model, so the authors recommend setting it on your own data.
Paper: https://t.co/mY9CuPl9sh
