Re-pin SONAR: exclude seedtts_testset (4548->3948)

#8
Speech Antispoofing Benchmarks by lab260 org

seedtts_testset clips are real Common Voice reference prompts, not synthesized speech; mislabeling them spoof inflated EER. New SHA eca7c72, n_trials 3948 (2274 bonafide / 1674 spoof).

Speech Antispoofing Benchmarks by lab260 org

Re-pin SONAR to seedtts-fixed revision (4548->3948).

korallll changed pull request status to merged

Sign up or log in to comment