TypeSafe AI Added to Citation Checker Tool
User added @typesafeai to their citation checker and praises it, free to try at PaperTrellis, tagged with Jev and AI in medicine. Link
Study on Jev's Probability Allocation in Multi-Answer Tasks
The post shares a study on model behavior in multi-answer tasks: models can rank correct options before distractors yet still assign low probabilities to some correct items. The paper evaluates Jev and other models on MetaTool, HotpotQA, and ECtHR.
多答案任务里，模型能把正确选项全排在干扰项前，却仍可能给其中一些正确项过低概率。 https://arxiv.org/abs/2610.29217 论文在 MetaTool、HotpotQA 和 ECtHR 上，读取 Jev、Liquid d1、GPT-6 Luna Decisions
· 0 likesOpen on X
More like this
JevBench: First Benchmark for Jev-Class Models Released
JevBench is the first benchmark for Jev-class models. Original Jev by TypeSafe AI leads at 75.3, with SemIf second at 74.6. Full results at benchmarkheaven.com/jev-models. Link
Jev Completes 100 Consecutive Trolley Problems
The post says someone entrusted their life to Jev and had it complete 100 consecutive trolley problems; in the first round, Jev chose to pull the lever with 99% probability, sacrificing 1 to save 5.
