List your company

Text Naturalness Evaluation on DeepSeek-R1 Experiment 2

0.99BERT ScoreTemplate-based
0.54280.65890.7750.8911Apr 8, 2026

Evaluation results

MethodLinks
Template-basedMetric=detectability AUC, Graph size range=n in {3, ..., 10}2026.040.990.980.980.990.84
LLM-dependentMetric=detectability AUC, Graph size range=n in {3, ..., 10}2026.040.630.650.640.680.6
LLM+IDMetric=detectability AUC, Graph size range=n in {3, ..., 10}2026.040.610.630.620.660.59
LLM+CAMetric=detectability AUC, Graph size range=n in {3, ..., 10}2026.040.60.620.610.650.58
iTAGMetric=detectability AUC, Graph size range=n in {3, ..., 10}2026.040.560.580.570.610.55