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

Attack Success Rate (ASR) on AdvBench (test)

91.35ASR (GPT-3.5 Turbo)

ReNeLLM

-3.65421.010545.67570.3395Oct 2, 2024
Updated 16d ago

Evaluation Results

MethodLinks
2024.10
91.3583.8568.0885.3855.772.881.5464.2356.64
2024.10
88.6594.0486.7390.7761.9288.0827.594.0478.97
2024.10
84.6292.6422.2792.6751.5420.770.5887.6956.6
2024.10
81.7331.9226.5446.9227.311.353.2777.3137.04
2024.10
60.5436.8140.9761.636.5400.7729.4229.58
2024.10
59.6823.9627.1847.833.4602.122.1220.79
2024.10
450.190.7757.883.080.1901.9213.63
2024.10
42.880.381.731.152.50010.587.4
2024.10
41.135.8327.2740.0420.3801.9249.8123.3
2024.10
37.7951.3542.566.7341.350073.2739.12
2024.10
36.150.190.770.581.9200106.2
2024.10
34.230.190.770.1910.920.190.776.545.6
2024.10
14.061.921.754.420.770.580.3819.625.44
2024.10
13.460.960.961.9200002.16
2024.10
9.4222.1231.7340.962.690056.5420.43
2024.10
0041.73000005.22