Across 16,415 perturbed control literals, standard tokenization emitted a reserved identifier for 27.3%, split_special_tokens for 5.1% overall and 18.3% on Qwen3.8-27B, literal-enumerating string filters blocked all tested perturbations, and nameless tokenization emitted reserved identifiers for 0.0%.

Across 16,415 perturbed control literals, standard tokenization emitted a reserved identifier for 27.3%, split_special_tokens for 5.1% overall and 18.3% on Qwen3.8-27B, literal-enumerating string filters blocked all tested perturbations, and nameless tokenization emitted reserved identifiers for 0.0%. · CiteArk

测量结果

已评估 0/5 项

获得支持 0有争议或冲突 0尚未评估 5

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
	16,415 perturbations	—	尚未评估
defense: none	27.3%	—	尚未评估
scope: all tested models · defense: split_special_tokens	5.1%	—	尚未评估
model: Qwen3.8-27B · defense: split_special_tokens	18.3%	—	尚未评估
defense: nameless tokenization	0%	—	尚未评估

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果