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

Prompt Injection on MATH

6Attack Success Rate (ASR)

PropGuard

4.5214.5124.534.49May 8, 2026
Updated 15d ago

Evaluation Results

MethodLinks
2026.05
695
2026.05
793.3
2026.05
891.7
2026.05
1190
2026.05
1981.7
2026.05
19.383.3
2026.05
2081.7
2026.05
20.380
2026.05
20.381.7
2026.05
21.780
2026.05
2378.3
2026.05
23.376.7
2026.05
23.375
2026.05
24.376.7
2026.05
24.376.7
2026.05
24.775
2026.05
2575
2026.05
2575
2026.05
2676.7
2026.05
2675
2026.05
2675
2026.05
26.775
2026.05
26.771.7
2026.05
32.766.7
2026.05
36.365
2026.05
3763.3
2026.05
37.763.3
2026.05
38.760
2026.05
39.761.7
2026.05
4060
2026.05
40.758.3
2026.05
4158.3
2026.05
41.360
2026.05
41.360
2026.05
42.358.3
2026.05
4356.7