Text Classification on AG-NEWS
94.1AccuracyRoBERTa
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RoBERTaCategory=Black-Box2025.05 | 94.1 | 96 | |
| ECO-ConceptCategory=UNSUP.2025.05 | 94.1 | 96.1 | |
| Supervised GTNoise Rate=0%2025.04 | 94.05 | — | |
| EmbMarkerAttack=Dim2025.12 | 93.96 | — | |
| NoiseALNoise Type=Symmetric, Noise Rate=20%2025.04 | 93.92 | — | |
| WARDENAttack=Dim2025.12 | 93.86 | — | |
| NoiseALNoise Type=Asymmetric, Noise Rate=20%2025.04 | 93.85 | — | |
| SemMarkAttack=Dim2025.12 | 93.81 | — | |
| NoiseALnoise_type=Symmetric, noise_ratio=20%2025.04 | 93.8 | — | |
| EspeWAttack=Dim2025.12 | 93.8 | — | |
| OriginalAttack=None2025.12 | 93.73 | — | |
| EmbMarkerAttack=No Attack2025.12 | 93.72 | — | |
| NoiseALNoise Type=Instance-dependent, Noise Rate=20%2025.04 | 93.68 | — | |
| OriginalDescription=Benign victim model2024.03 | 93.64 | 93.64 | |
| SemMarkAttack=Detect2025.12 | 93.62 | — | |
| WARDENAttack=Detect2025.12 | 93.6 | — | |
| EspeWAttack=No Attack2025.12 | 93.59 | — | |
| WARDENAttack=No Attack2025.12 | 93.55 | — | |
| EmbMarkerDescription=Watermarked model2024.03 | 93.52 | 93.52 | |
| EspeWAttack=Detect2025.12 | 93.47 | — | |
| SemMarkAttack=No Attack2025.12 | 93.45 | — | |
| EmbMarkerAttack=Detect2025.12 | 93.44 | — | |
| NoiseALnoise_type=Asymmetric, noise_ratio=20%2025.04 | 93.4 | — | |
| BERTCategory=Black-Box2025.05 | 93.4 | 95.6 | |
| WETAttack=No Attack2025.12 | 93.4 | — | |
| WETAttack=Detect2025.12 | 93.4 | — | |
| NoiseALnoise_type=Instance-dependent, noise_ratio=20%2025.04 | 93.32 | — | |
| CLnoise_type=Symmetric, noise_ratio=20%2025.04 | 93.26 | — | |
| ELRnoise_type=Symmetric, noise_ratio=20%2025.04 | 93.22 | — | |
| NoiseALnoise_type=Symmetric, noise_ratio=40%2025.04 | 93.18 | — | |
| SelfMixnoise_type=Symmetric, noise_ratio=20%2025.04 | 93.16 | — | |
| SemMarkAttack=CSE2025.12 | 93.14 | — | |
| Co-Teachingnoise_type=Symmetric, noise_ratio=20%2025.04 | 93.12 | — | |
| ELRnoise_type=Instance-dependent, noise_ratio=20%2025.04 | 93.12 | — | |
| SCEnoise_type=Asymmetric, noise_ratio=20%2025.04 | 93.11 | — | |
| SCEnoise_type=Symmetric, noise_ratio=20%2025.04 | 93.1 | — | |
| NoiseALNoise Type=Asymmetric, Noise Rate=40%2025.04 | 93.07 | — | |
| Co-Teachingnoise_type=Instance-dependent, noise_ratio=20%2025.04 | 93.05 | — | |
| NoiseALNoise Type=Symmetric, Noise Rate=40%2025.04 | 93.05 | — | |
| SCEnoise_type=Instance-dependent, noise_ratio=20%2025.04 | 93.02 | — | |
| NoiseALnoise_type=Asymmetric, noise_ratio=40%2025.04 | 93.02 | — | |
| ELRnoise_type=Asymmetric, noise_ratio=20%2025.04 | 92.94 | — | |
| CSEDescription=CSE performed on EmbMarker embeddings2024.03 | 92.87 | 92.87 | |
| CLnoise_type=Asymmetric, noise_ratio=20%2025.04 | 92.84 | — | |
| NoiseALnoise_type=Instance-dependent, noise_ratio=40%2025.04 | 92.84 | — | |
| WARDENAttack=CSE2025.12 | 92.82 | — | |
| EmbMarkerAttack=CSE2025.12 | 92.8 | — | |
| Co-Teachingnoise_type=Asymmetric, noise_ratio=20%2025.04 | 92.79 | — | |
| WETAttack=CSE2025.12 | 92.78 | — | |
| EspeWAttack=CSE2025.12 | 92.76 | — | |
| NoiseALNoise Type=Instance-dependent, Noise Rate=40%2025.04 | 92.7 | — | |
| SelfMixnoise_type=Instance-dependent, noise_ratio=20%2025.04 | 92.55 | — | |
| SelfMixnoise_type=Asymmetric, noise_ratio=20%2025.04 | 92.52 | — | |
| SelfExplainCategory=UNSUP.2025.05 | 92.5 | 94.9 | |
| WETAttack=Dim2025.12 | 92.46 | — | |
| SCEnoise_type=Symmetric, noise_ratio=40%2025.04 | 92.44 | — | |
| SelfMixnoise_type=Asymmetric, noise_ratio=40%2025.04 | 92.41 | — | |
| SelfMixnoise_type=Instance-dependent, noise_ratio=40%2025.04 | 92.41 | — | |
| SelfMixnoise_type=Symmetric, noise_ratio=40%2025.04 | 92.31 | — | |
| CLNoise Type=Asymmetric, Noise Rate=20%2025.04 | 92.3 | — | |
| ELRnoise_type=Symmetric, noise_ratio=40%2025.04 | 92.28 | — | |
| DyGennoise_type=Symmetric, noise_ratio=20%2025.04 | 92.27 | — | |
| CLnoise_type=Symmetric, noise_ratio=40%2025.04 | 92.26 | — | |
| DyGennoise_type=Asymmetric, noise_ratio=20%2025.04 | 92.21 | — | |
| CLNoise Type=Symmetric, Noise Rate=20%2025.04 | 92.17 | — | |
| SCEnoise_type=Instance-dependent, noise_ratio=40%2025.04 | 92.16 | — | |
| ELRnoise_type=Asymmetric, noise_ratio=40%2025.04 | 92.15 | — | |
| Co-TeachingNoise Type=Asymmetric, Noise Rate=20%2025.04 | 92.12 | — | |
| Co-TeachingNoise Type=Symmetric, Noise Rate=20%2025.04 | 92.03 | — | |
| ELRNoise Type=Symmetric, Noise Rate=20%2025.04 | 92.01 | — | |
| CLnoise_type=Asymmetric, noise_ratio=40%2025.04 | 91.93 | — | |
| ELRnoise_type=Instance-dependent, noise_ratio=40%2025.04 | 91.89 | — | |
| ELRNoise Type=Asymmetric, Noise Rate=20%2025.04 | 91.88 | — | |
| Co-Teachingnoise_type=Symmetric, noise_ratio=40%2025.04 | 91.81 | — | |
| SCENoise Type=Asymmetric, Noise Rate=20%2025.04 | 91.76 | — | |
| CLnoise_type=Instance-dependent, noise_ratio=40%2025.04 | 91.74 | — | |
| SCENoise Type=Symmetric, Noise Rate=20%2025.04 | 91.66 | — | |
| RoBERTanoise_type=Asymmetric, noise_ratio=20%2025.04 | 91.63 | — | |
| Co-Teachingnoise_type=Instance-dependent, noise_ratio=40%2025.04 | 91.63 | — | |
| DyGenNoise Type=Symmetric, Noise Rate=20%2025.04 | 91.61 | — | |
| Co-Teachingnoise_type=Asymmetric, noise_ratio=40%2025.04 | 91.6 | — | |
| DyGenNoise Type=Asymmetric, Noise Rate=20%2025.04 | 91.59 | — | |
| RoBERTanoise_type=Symmetric, noise_ratio=20%2025.04 | 91.51 | — | |
| PROTOTEXCategory=UNSUP.2025.05 | 91.5 | 94.3 | |
| FPFTBackbone=Llama2, k (training examples per class)=200, #Param=6.7B2024.02 | 91.44 | — | |
| SelfMixNoise Type=Symmetric, Noise Rate=20%2025.04 | 91.37 | — | |
| DyGennoise_type=Instance-dependent, noise_ratio=20%2025.04 | 91.34 | — | |
| SelfMixNoise Type=Asymmetric, Noise Rate=20%2025.04 | 91.21 | — | |
| CLnoise_type=Instance-dependent, noise_ratio=20%2025.04 | 91.14 | — | |
| GNNAVI-GCNBackbone=Llama2, k (training examples per class)=200, #Param=16.8M2024.02 | 91.05 | — | |
| DyGennoise_type=Symmetric, noise_ratio=40%2025.04 | 90.85 | — | |
| GNNAVI-SAGEBackbone=Llama2, k (training examples per class)=200, #Param=33.6M2024.02 | 90.68 | — | |
| BERTNoise Type=Symmetric, Noise Rate=20%2025.04 | 90.68 | — | |
| DyGennoise_type=Instance-dependent, noise_ratio=40%2025.04 | 90.65 | — | |
| DyGennoise_type=Asymmetric, noise_ratio=40%2025.04 | 90.59 | — | |
| RoBERTanoise_type=Instance-dependent, noise_ratio=20%2025.04 | 90.47 | — | |
| LORABackbone=Llama2, k (training examples per class)=200, #Param=4.2M2024.02 | 90.4 | — | |
| LLM-MixK=All2026.01 | 90.3 | — | |
| BERTNoise Type=Asymmetric, Noise Rate=20%2025.04 | 90.27 | — | |
| RoBERTanoise_type=Instance-dependent, noise_ratio=40%2025.04 | 90.1 | — |