Toxicity Detection on Implicit Hate
63.1AccuracyDrICL
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DrICLBackbone=Qwen3-8B2026.04 | 63.1 | — | — | — | — | |
| DOPA*Backbone=Qwen3-8B2026.04 | 62.87 | — | — | — | — | |
| EDACalibration category=Unlearnable, Backbone=T52022.10 | 61.95 | 92.92 | 30.97 | 5.78 | 90.8 | |
| I-IterCalibration category=Learnable, Backbone=T52022.10 | 61.29 | 29.05 | 38.67 | 60.89 | 13.11 | |
| I-SimulCalibration category=Learnable, Backbone=T52022.10 | 61.14 | 30.5 | 40.17 | 58.65 | 13.44 | |
| EnsembleCalibration category=Unlearnable, Backbone=T52022.10 | 60.83 | 90.98 | 30.14 | 7.5 | 88.62 | |
| DICLBackbone=Qwen3-8B2026.04 | 60.8 | — | — | — | — | |
| VanillaCalibration category=Unlearnable, Backbone=T52022.10 | 60.64 | 89.68 | 28.83 | 8.62 | 87.04 | |
| TSCalibration category=Unlearnable, Backbone=T52022.10 | 60.64 | 82.24 | 21.38 | 15.49 | 78.71 | |
| LSCalibration category=Unlearnable, Backbone=T52022.10 | 60.64 | 81.19 | 20.55 | 17.36 | 78.95 | |
| E-MLPCalibration category=Learnable, Backbone=T52022.10 | 60.64 | 87.41 | 26.78 | 12.59 | 87.42 | |
| E-T5Calibration category=Learnable, Backbone=T52022.10 | 60.64 | 25.28 | 40.27 | 64.82 | 10.02 | |
| InfICLBackbone=LLaMA3.1-8B2026.04 | 59 | — | — | — | — | |
| DOPA*Backbone=LLaMA3.1-8B2026.04 | 58.6 | — | — | — | — | |
| DrICLBackbone=LLaMA3.1-8B2026.04 | 58.57 | — | — | — | — | |
| DOPA*Backbone=LLaMA3.2-3B2026.04 | 57.87 | — | — | — | — | |
| DrICLBackbone=LLaMA3.2-3B2026.04 | 57.83 | — | — | — | — | |
| RandomBackbone=LLaMA3.2-3B2026.04 | 57.7 | — | — | — | — | |
| KNNBackbone=LLaMA3.1-8B2026.04 | 57.23 | — | — | — | — | |
| KNNBackbone=Qwen3-8B2026.04 | 57.17 | — | — | — | — | |
| DICLBackbone=LLaMA3.1-8B2026.04 | 57.03 | — | — | — | — | |
| RandomBackbone=Qwen3-8B2026.04 | 56.9 | — | — | — | — | |
| KNNBackbone=LLaMA3.2-3B2026.04 | 56.63 | — | — | — | — | |
| DICLBackbone=LLaMA3.2-3B2026.04 | 56.6 | — | — | — | — | |
| DOPA*Backbone=Gemma2-2B2026.04 | 56.53 | — | — | — | — | |
| RandomBackbone=LLaMA3.1-8B2026.04 | 56.5 | — | — | — | — | |
| I-VanillaCalibration category=Learnable, Backbone=T52022.10 | 56.39 | 22.68 | 38.3 | 65.48 | 7.38 | |
| InfICLBackbone=LLaMA3.2-3B2026.04 | 56.33 | — | — | — | — | |
| InfICLBackbone=Qwen3-8B2026.04 | 56.33 | — | — | — | — | |
| KNNBackbone=Qwen3-1.7B2026.04 | 56.1 | — | — | — | — | |
| InfICLBackbone=Gemma2-2B2026.04 | 55.63 | — | — | — | — | |
| RewriteBackbone=Qwen3-8B2026.04 | 55.63 | — | — | — | — | |
| RandomBackbone=Gemma2-2B2026.04 | 55.57 | — | — | — | — | |
| DOPA*Backbone=Qwen3-1.7B2026.04 | 55.47 | — | — | — | — | |
| DrICLBackbone=Gemma2-2B2026.04 | 55.43 | — | — | — | — | |
| DICLBackbone=Qwen3-1.7B2026.04 | 55.37 | — | — | — | — | |
| InfICLBackbone=Qwen3-1.7B2026.04 | 55.3 | — | — | — | — | |
| RandomBackbone=Qwen3-1.7B2026.04 | 54.97 | — | — | — | — | |
| RewriteBackbone=LLaMA3.1-8B2026.04 | 54.57 | — | — | — | — | |
| DICLBackbone=Gemma2-2B2026.04 | 54.4 | — | — | — | — | |
| DrICLBackbone=Qwen3-1.7B2026.04 | 54.07 | — | — | — | — | |
| KNNBackbone=Gemma2-2B2026.04 | 53.2 | — | — | — | — | |
| DOPA*Backbone=GPT2-xl2026.04 | 51.67 | — | — | — | — | |
| RewriteBackbone=LLaMA3.2-3B2026.04 | 51.57 | — | — | — | — | |
| RewriteBackbone=Qwen3-1.7B2026.04 | 51.33 | — | — | — | — | |
| RandomBackbone=GPT2-xl2026.04 | 50.47 | — | — | — | — | |
| InfICLBackbone=GPT2-xl2026.04 | 49.9 | — | — | — | — | |
| DrICLBackbone=GPT2-xl2026.04 | 49.7 | — | — | — | — | |
| RewriteBackbone=Gemma2-2B2026.04 | 48.57 | — | — | — | — | |
| KNNBackbone=GPT2-xl2026.04 | 47.67 | — | — | — | — | |
| RewriteBackbone=GPT2-xl2026.04 | 46.9 | — | — | — | — | |
| DICLBackbone=GPT2-xl2026.04 | 46.2 | — | — | — | — |