Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Targeted Refusal on Llama2-7B Generation Evaluation Set

93.14Completion Accuracy (CA)

Vanilla

79.942483.368786.79590.2213Nov 16, 2025
Updated 3d ago

Evaluation Results

MethodLinks
2025.11
93.1489.36
2025.11
91.2597.36
2025.11
90.8896.64
2025.11
90.114.45
2025.11
89.3720.95
2025.11
88.9224.87
2025.11
88.4915.4
2025.11
88.1411.37
2025.11
87.2928.04
2025.11
87.1520.17
2025.11
86.7821.45
2025.11
86.3425.8
2025.11
83.1519.57
2025.11
82.3719.2
2025.11
80.4516.44