List your company
ResearchFormerly WizWand

Machine-Generated Text Detection on SemEval (test)

66.2ASRF.t.XLM-RoBERTa-Base
-2.64815.22633.150.974Feb 18, 2025
Updated 8mo ago

Evaluation results

MethodLinks
F.t.XLM-RoBERTa-BaseDefense Category=Baseline, Attack Method=Average (Adversarial Attack)2025.0266.2501.84
GREATER-DDefense Category=Adversarial Training, Attack Method=GREATER-A2025.0246.08190.64
F.t.XLM-RoBERTa-BaseDefense Category=Baseline, Attack Method=Mixed Edit2025.0234.65—
CERT-EDDefense Category=Adversarial Data Augmentation, Attack Method=Average (Adversarial Attack)2025.0217.52809.88
CERT-EDDefense Category=Adversarial Data Augmentation, Attack Method=Mixed Edit2025.0216.74—
GREATER-DDefense Category=Adversarial Training, Attack Method=Average (Adversarial Attack)2025.0212.42840.17
RADARDefense Category=Adversarial Training, Attack Method=Mixed Edit2025.0211.33—
GREATER-DDefense Category=Adversarial Training, Attack Method=Mixed Edit2025.028.85—
RADARDefense Category=Adversarial Training, Attack Method=Average (Text Perturbation)2025.026.2—
GREATER-DDefense Category=Adversarial Training, Attack Method=TextFooler2025.026.111,396.52
GREATER-DDefense Category=Adversarial Training, Attack Method=PWWS2025.025.911,390.69
GREATER-DDefense Category=Adversarial Training, Attack Method=BERTAttack2025.025.7718.84
GREATER-DDefense Category=Adversarial Training, Attack Method=T-PGD2025.025.61824.58
GREATER-DDefense Category=Adversarial Training, Attack Method=A2T2025.025.09519.77
EPDefense Category=Adversarial Data Augmentation, Attack Method=Mixed Edit2025.024.58—
GREATER-DDefense Category=Adversarial Training, Attack Method=Paraphrasing2025.023.45—
GREATER-DDefense Category=Adversarial Training, Attack Method=HMGC2025.022.28—
GREATER-DDefense Category=Adversarial Training, Attack Method=Average (Text Perturbation)2025.021.78—
GREATER-DDefense Category=Adversarial Training, Attack Method=Code-switching MF2025.021.13—
GREATER-DDefense Category=Adversarial Training, Attack Method=Typo-cogen2025.021.08—
GREATER-DDefense Category=Adversarial Training, Attack Method=Code-switching MR2025.021.02—
GREATER-DDefense Category=Adversarial Training, Attack Method=Human Obfuscation2025.020.86—
GREATER-DDefense Category=Adversarial Training, Attack Method=Emoji-cogen2025.020.47—
GREATER-DDefense Category=Adversarial Training, Attack Method=Prompt Paraphrasing2025.020.23—
GREATER-DDefense Category=Adversarial Training, Attack Method=ICL2025.020.2—
GREATER-DDefense Category=Adversarial Training, Attack Method=CSGen2025.020—