Datasets:

Modalities:
Audio
Text
Formats:
parquet
Languages:
English
ArXiv:
License:

Add M1 submission

#1
by mbocanu - opened
No description provided.
Speech Antispoofing Benchmarks by lab260 org

speech-spoof-bench ci verify-pr — ❌ failures present

Submission Schema sha256 EER match Notes
submissions/m1.yaml ✅ ❌ ❌ reproduce --scoring exited non-zero (see job log)

🤖 view CI run

The first CI run failed because the submission lacked the calibration block that srr_complement needs. It is now added (DeepVoice threshold 2.0103, as computed by the toolkit). Could you re-run the check? Thanks.

Speech Antispoofing Benchmarks by lab260 org

Verified: verify-pr CI green + maintainer reproduce --scoring matched (Δ 0). Thanks for the submission! Merging.

korallll changed pull request status to merged

Sign up or log in to comment