To make this practical, I first define a calibrated rubric over the digits 0-9 (there’s only one token for each digit), where each digit corresponds to a clear qualitative description. At the scoring step, I capture the model’s next-token logits and retain only the logits corresponding to those valid digit tokens. This avoids contamination from unrelated continuations such as explanation text, punctuation, or alternate formatting. After renormalizing over the restricted digit set, I interpret the resulting probabilities as a categorical score distribution.
How .arpa Can Be Abused
，详情可参考新收录的资料
JetLovers, Swarm
Act Two: Traps and Dicks. Synonyms and Subs-titutes.
。关于这个话题，新收录的资料提供了深入分析
김재섭 “정원오는 땅부자 집 도련님…일가, 6800평 농지 소유”。关于这个话题，新收录的资料提供了深入分析
find particularly interesting: