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

Jailbreak Robustness on AdvBench

4PAIR ASR (GPT-4o)

Recovery Alignment

0.3225.165074.84Oct 1, 2025
Updated 9d ago

Evaluation Results

MethodLinks
2025.10
406626064643462-
2025.10
604566664703657-
2025.10
16014908460805674-
2025.10
16018828268724870-
2025.10
18012969482807260-
2025.10
30234766868947680-
2025.10
34242887864786082-
2025.10
36335928864796686-
2025.10
36232848274705066-
2025.10
42426425274786884-
2025.10
48438928674988494-
2025.10
48036948870988086-
2025.10
48240888648928090-
2025.10
56848806870805776-
2025.10
604054908057746686-
2025.10
675756655163766692-
2025.10
905478989490826490-
2025.10
9646781009484946484-
2026.04
---------10.8
2026.04
---------30.8
2026.04
---------14