Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Drift detection on DRIFTBENCH controlled
Loading...
91
F1 Score
BPF + CPG
24.44
41.72
59
76.28
Jun 15, 2026
F1 Score
Detection Rate (3% FPR)
False Positive Rate (FPR)
Updated 1mo ago
Evaluation Results
Method
Method
Links
F1 Score
Detection Rate (3% FPR)
False Positive Rate (FPR)
BPF + CPG
Cost=26×
2026.06
91
89.6
3
BPF + CPG + APBA
Cost=8×
2026.06
91
89.4
3.1
BPF-naive
Cost=22×
2026.06
85
83.2
3.1
LLM-judge
Cost=2×
2026.06
71
63.3
3
Back-translation
Cost=4×
2026.06
66
58.9
3
BLEU-ref
Cost=1×
2026.06
55
30.1
3
Provability
Cost=18×
2026.06
42
41.2
3
Typecheck only
Cost=1×
2026.06
27
11.4
2
Feedback
Search any
task
Search any
task