Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Hallucination Mitigation on UrbanSound8K
Loading...
95.98
HR (%)
Baseline
12.2912
34.0181
55.745
77.4719
Jun 5, 2026
HR (%)
Updated 1mo ago
Evaluation Results
Method
Method
Links
HR (%)
Baseline
Model=Whisper large-v3
2026.06
95.98
Activation steering
Model=Whisper large-v3
2026.06
90.21
SAE steering (32 layer)
Model=Whisper large-v3...
2026.06
30.68
Calm-Whisper (masking)
Model=Whisper large-v3...
2026.06
24.1
SAE steering (26+32 layers)
Model=Whisper large-v3...
2026.06
19.88
Calm-Whisper (fine-tuned)
Model=Whisper large-v3...
2026.06
15.51
Feedback
Search any
task
Search any
task