Jevals Data: Independent benchmark for TypeSafe Jev
Independent benchmark data comparing TypeSafe AI's Jev (System One model) against LLMs on decision score, accuracy, calibration, cost, and latency for typed tasks.
A sealed-bid auction simulator that uses TypeSafe Jev's typed probability probes to estimate value distributions, with calibration and regret evaluation.
该仓库是一个密封拍卖模拟器,明确调用 TypeSafe Jev API 的概率探针来估计价值分布,并包含在线运行、校准与后悔值评估,属于对 Jev 模型的真实集成与实验研究。