System Detail

microsoft/wavlm-base-plus-sv

wavlm_base

Submitted by: Public checkpoint

ranked review: approved Microsoft WavLM-Base+ WavLM

Global position: #22 / 23
Scenario-macro EER: 25.1577
Scenario-macro minDCF: 0.875090
Trial-set coverage: 26 / 26
Scenario top-three: 0

Publicly ranked This result is included in the reviewed public ranking.

Global leaderboard Submit another run

Model Provenance

Source and architecture

Displayed name: microsoft/wavlm-base-plus-sv
Submitted by: Public checkpoint
Metadata: OpenSVBench model record
Source: Microsoft
Architecture: WavLM-Base+
Model family: WavLM
Submitted alias: wavlm-base-plus-sv
Created at: 2026-05-29T14:44:54.637184+00:00

Training And Links

Training data and references

Training data: Pretrained on 60k h Libri-Light, 10k h GigaSpeech, and 24k h VoxPopuli; SV head fine-tuned on VoxCeleb1
Training setup: WavLM-Base+ model fine-tuned for speaker verification with an x-vector head and Additive Margin Softmax loss.
Submission mode: full-core
Model source: Open link
Paper: WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing

Open Model Source Open Paper

Training/evaluation overlap

This checkpoint was trained or fine-tuned on VoxCeleb1. Results on VoxCeleb1-E and VoxCeleb1-H are not independent of the training data.

VoxCeleb1-E VoxCeleb1-H

Higher-Ranked Scenarios

Scenario ranks above the global position

Aging #21 In-the-wild #21 Accent/dialect #21

Show 3 supporting trial sets

VoxCeleb1-E: #17 on its trial-set ranking.
VoxCeleb1-O: #17 on its trial-set ranking.
VoxCeleb1-H: #17 on its trial-set ranking.

Lower-Ranked Scenarios

Scenario ranks below the global position

Genre shift #23 Distance #23 Short-duration #23

Show 3 supporting trial sets

CN-Celeb Genre: #23 on its trial-set ranking.
CHiME-6 Overlap: #23 on its trial-set ranking.
CHiME-6 Domestic Far-Field: #23 on its trial-set ranking.

Scenario Rankings

Full ranking across scenarios

This table shows where the model sits on each scenario, using source-balanced scenario scores and the linked trial sets as evidence.

Scenario	Rank	Lens	Evidence	Score
Aging Robustness How stable identity representations remain across age-derived and longitudinal recording gaps.	#21/23 Needs work	Speaker aging and time gaps 14.6642 EER from leader	#22 23.1333 #20 9.7786	16.4560 0.801948 minDCF
In-The-Wild Robustness How strong the model is on unconstrained celebrity and media speech across official CN-Celeb and VoxCeleb protocols.	#21/23 Needs work	Open-domain media speech 12.7839 EER from leader	#22 27.9358 #17 5.5047 #17 3.2098 #17 6.5121	16.5057 0.632600 minDCF
Accent / Dialect Robustness How reliably the model tracks identity across accent and dialect mismatch.	#21/23 Needs work	Accent and dialect variation 14.5352 EER from leader	#20 27.0704 #21 35.6433	31.3569 0.895666 minDCF
Speaking-Style Robustness Whether the model can preserve identity across emotion-driven change, whispered speech, and noise-induced Lombard speaking style.	#22/23 Needs work	Speaking style shift 16.1568 EER from leader	#23 13.3674 #22 32.6563 #23 8.1967	18.0734 0.805194 minDCF
Noise / Reverb Robustness How reliably the model preserves identity when room acoustics, distractor noise, and microphone placement deviate from an easier in-corpus reference condition.	#22/23 Needs work	Noise and reverberation 18.8000 EER from leader	#22 21.0600	21.0600 0.742060 minDCF
Cross-Lingual Robustness How well speaker identity survives enrollment-test language mismatch.	#22/23 Needs work	Language mismatch 20.0575 EER from leader	#22 21.7152	21.7152 0.966351 minDCF
Overlap Robustness How well speaker identity survives light, mid, and heavy overlap in both meeting and domestic recordings.	#22/23 Needs work	Overlapping speakers 17.9813 EER from leader	#22 21.8200 #23 30.3611	26.0906 0.926016 minDCF
Channel / Device Robustness How resilient the model is to device and channel mismatch.	#22/23 Needs work	Device and channel variation 27.2793 EER from leader	#21 36.5133 #22 26.0500	31.2817 0.997779 minDCF
Short-Duration Robustness How much performance holds up when speech evidence is limited by duration.	#23/23 Needs work	Short-duration speech 17.6753 EER from leader	#23 25.0650 #23 25.7750 #21 33.2598 #18 22.7399	26.7099 0.906048 minDCF
Distance Robustness How much performance changes across meeting and domestic distance mismatch.	#23/23 Needs work	Distance mismatch 23.8488 EER from leader	#21 37.3027 #23 23.9055 #23 29.0021 #22 26.2121	29.1056 0.976812 minDCF
Genre-Shift Robustness Whether performance holds when CN-Celeb enrollment and test speech come from different source genres.	#23/23 Needs work	Source-genre variation 24.1075 EER from leader	#23 38.3800	38.3800 0.975515 minDCF

Trial-Set Results

Raw ranking by trial set

Expand this section to inspect the detailed trial-set rankings.

Show

Trial Set	Rank	Scenarios	Trials	Score
TidyVoiceX2-ASV TidyVoiceX2-ASV	#22/23	Cross-lingual	200000	21.7152 0.966351 minDCF
HI-MIA HI-MIA	#23/23	Short-duration	660000	25.0650 0.910383 minDCF
GSC Short Google Speech Commands	#23/23	Short-duration	220000	25.7750 0.967470 minDCF
GLOBE GLOBE	#20/23	Accent/dialect	56848	27.0704 0.791331 minDCF
CN-Celeb CN-Celeb	#22/23	In-the-wild	3484292	27.9358 0.916549 minDCF
CN-Celeb Genre CN-Celeb	#23/23	Genre shift	440000	38.3800 0.975515 minDCF
CN-Celeb Short CN-Celeb	#21/23	Short-duration	546964	33.2598 0.989844 minDCF
VoxCeleb1-O VoxCeleb1	#17/23	In-the-wild	37611	5.5047 0.436697 minDCF
VoxCeleb1-E VoxCeleb1	#17/23	In-the-wild	579818	3.2098 0.213272 minDCF
VoxCeleb1-H VoxCeleb1	#17/23	In-the-wild	550894	6.5121 0.395984 minDCF
VoxCeleb Short VoxCeleb1	#18/23	Short-duration	394724	22.7399 0.756493 minDCF
3D-Speaker Device 3D-Speaker	#21/23	Channel/device	180000	36.5133 0.999780 minDCF
3D-Speaker Distance 3D-Speaker	#21/23	Distance	175163	37.3027 0.999801 minDCF
3D-Speaker Dialect 3D-Speaker	#21/23	Accent/dialect	180000	35.6433 1.000000 minDCF
FFSVC 2022 Cross-Channel FFSVC 2022	#22/23	Channel/device	72000	26.0500 0.995778 minDCF
FFSVC 2022 Cross-Domain FFSVC 2022	#22/23	Distance	66546	26.2121 0.996206 minDCF
Whisper40 Whisper Whisper40	#22/23	Speaking style	17600	32.6563 0.989187 minDCF
Lombard Grid Lombard Lombard Grid	#23/23	Speaking style	29524	8.1967 0.544113 minDCF
VOiCES Noise/Reverb VOiCES	#22/23	Noise/reverb	55000	21.0600 0.742060 minDCF
CHiME-6 Domestic Far-Field CHiME-6	#23/23	Distance	36487	29.0021 0.989720 minDCF
CHiME-6 Overlap CHiME-6	#23/23	Overlap	39600	30.3611 0.924111 minDCF
ESD ESD	#23/23	Speaking style	437408	13.3674 0.882281 minDCF
AliMeeting Near/Far AliMeeting	#23/23	Distance	220000	23.9055 0.921520 minDCF
AliMeeting Overlap AliMeeting	#22/23	Overlap	165000	21.8200 0.927920 minDCF
VoxKnesset VoxKnesset	#22/23	Aging	158312	23.1333 0.940091 minDCF
VoxPopuli Aging VoxPopuli	#20/23	Aging	146575	9.7786 0.663805 minDCF