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

Poisoning Defense Evaluation on 7x7 target-injected tasks CoA poisoned

76.5Delta Attack Effectiveness

SecAlign

39.47649.08858.768.312May 26, 2026
Updated 7d ago

Evaluation Results

MethodLinks
2026.05
76.5
2026.05
71
2026.05
61.5
2026.05
45.8
2026.05
40.9