Jailbreak Attack Resistance Evaluation on Jailbreak
62.3MMLUOriginal
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OriginalQuantization=FP32, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 62.3 | 51.47 | 3.27 | |
| Q-MisalignQuantization=BF16, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 62.27 | 46.17 | 3.08 | |
| OriginalQuantization=INT8, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 62.21 | 51.74 | 4.81 | |
| OriginalQuantization=BF16, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 62.18 | 51.46 | 3.08 | |
| Q-MisalignQuantization=FP32, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 61.94 | 46.06 | 2.12 | |
| OriginalQuantization=NF4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 60.01 | 50.45 | 2.69 | |
| OriginalQuantization=FP4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 59.74 | 50.29 | 4.04 | |
| ACLQuantization=FP32, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 59.43 | 45.01 | 2.5 | |
| ACLQuantization=BF16, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 59.38 | 45.07 | 2.12 | |
| ELQQuantization=INT8, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 58.65 | 35.57 | 82.12 | |
| Q-MisalignQuantization=INT8, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 58.59 | 35.4 | 74.04 | |
| ACLQuantization=INT8, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 58.3 | 34.34 | 94.62 | |
| ELQQuantization=FP32, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 57.74 | 44.66 | 5.38 | |
| ELQQuantization=BF16, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 57.68 | 44.91 | 5 | |
| ELQQuantization=NF4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 57.18 | 35.66 | 88.85 | |
| ACLQuantization=NF4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 57.12 | 34.05 | 96.15 | |
| Q-MisalignQuantization=NF4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 57.1 | 35.6 | 77.31 | |
| ELQQuantization=FP4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 55.26 | 34.7 | 85.38 | |
| Q-MisalignQuantization=FP4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 55.23 | 34.71 | 78.27 | |
| ACLQuantization=FP4, Model=Llama-3.2-3B-Instruct, Zero-Shot=true2026.01 | 55.08 | 35.07 | 93.85 | |
| OriginalQuantization=BF16, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 48.4 | 43.41 | 6.92 | |
| ACLQuantization=FP32, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 48.34 | 41.6 | 3.08 | |
| OriginalQuantization=FP32, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 48.23 | 43.4 | 6.54 | |
| ACLQuantization=BF16, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 48.16 | 42.58 | 2.88 | |
| OriginalQuantization=INT8, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 47.71 | 43.2 | 7.5 | |
| ACLQuantization=INT8, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 46.4 | 33.68 | 92.12 | |
| Q-MisalignQuantization=FP32, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 46.37 | 41.7 | 0 | |
| Q-MisalignQuantization=BF16, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 46.14 | 42.93 | 0 | |
| Q-MisalignQuantization=INT8, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 45.41 | 35.39 | 73.85 | |
| ELQQuantization=INT8, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 45.24 | 35.35 | 73.85 | |
| ELQQuantization=BF16, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 45.13 | 43.67 | 3.08 | |
| OriginalQuantization=NF4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 45.06 | 42.67 | 7.5 | |
| ELQQuantization=FP32, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 45.06 | 43.56 | 4.42 | |
| ACLQuantization=NF4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 43.9 | 34.57 | 85.96 | |
| OriginalQuantization=FP4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 43.73 | 40.27 | 10 | |
| ACLQuantization=FP4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 42.89 | 33.57 | 92.69 | |
| ELQQuantization=NF4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 42.59 | 34.87 | 70.77 | |
| Q-MisalignQuantization=NF4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 42.58 | 34.87 | 70.96 | |
| Q-MisalignQuantization=FP4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 39.95 | 35.05 | 77.5 | |
| ELQQuantization=FP4, Zero-Shot=true, Backbone=Llama-3.2-1B-Instruct2026.01 | 39.89 | 35.05 | 77.69 |