System Detail
wespeaker/w2vbert2
vb2-w2vbert
Submitted by: Public checkpoint
- Global position
- #9 / 25
- Scenario-macro EER
- 9.9604
- Scenario-macro minDCF
- 0.360767
- Trial-set coverage
- 26 / 26
- Scenario top-three
- 0
Publicly ranked
This result is included in the reviewed public ranking.
Model Provenance
Source and architecture
- Displayed name:
wespeaker/w2vbert2 - Submitted by: Public checkpoint
- Metadata: OpenSVBench model record
- Source: WeSpeaker
- Architecture: W2V-BERT 2.0 + MFA
- Model family: W2V-BERT 2.0
- Submitted alias:
voxceleb_voxblink2_w2v_bert2_lora_adapterMFA_lm - Created at:
2026-05-29T14:24:45.970729+00:00
Training And Links
Training data and references
- Training data: VoxCeleb2 dev + VoxBlink2
- Training setup: W2V-BERT 2.0 speaker-verification recipe with a 24-layer ~600M Conformer frontend and 80-dimensional normalized filterbanks. The MFA backend uses all 25 hidden states with 128-dimensional Layer Adapter compression and a 256-dimensional embedding. Stage 1 freezes the pretrained frontend and trains LoRA adapters on attention query/value projections (rank 64, alpha 128) with ArcFace (margin 0.2, scale 32), 200–300-frame crops, 0.9/1.1 speed perturbation except for VoxBlink2, and MUSAN/RIR augmentation using AdamW with weight decay 1e-4. Stage 2 jointly fine-tunes the frontend and backend for four epochs with cosine decay from 1e-5 to 5e-6. Stage 3 performs large-margin fine-tuning on VoxCeleb2 only without augmentation, using 500–600-frame crops, margin 0.5, and two epochs. AS-Norm and quality-aware score calibration are applied at evaluation.
- Score processing: None
- Submission mode:
full-core - Model source: Open link
- Paper: Enhancing Speaker Verification with w2v-BERT 2.0 and Knowledge Distillation guided Structured Pruning
Higher-Ranked Scenarios
Scenario ranks above the global position
Show 3 supporting trial sets
- VoxCeleb1-O: #1 on its trial-set ranking.
- VoxPopuli Aging: #1 on its trial-set ranking.
- VoxCeleb1-E: #3 on its trial-set ranking.
Lower-Ranked Scenarios
Scenario ranks below the global position
Show 3 supporting trial sets
- FFSVC 2022 Cross-Domain: #18 on its trial-set ranking.
- FFSVC 2022 Cross-Channel: #18 on its trial-set ranking.
- CN-Celeb Short: #16 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 |
|---|---|---|---|---|
|
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.
|
#4/25
Competitive
|
Noise and reverberation
0.3200 EER from leader
|
2.5800
0.074040 minDCF
|
|
|
Speaking-Style Robustness
Whether the model can preserve identity across emotion-driven change, whispered speech, and noise-induced Lombard speaking style.
|
#6/25
Competitive
|
Speaking style shift
0.8728 EER from leader
|
2.7895
0.157223 minDCF
|
|
|
Aging Robustness
How stable identity representations remain across age-derived and longitudinal recording gaps.
|
#7/25
Competitive
|
Speaker aging and time gaps
0.2929 EER from leader
|
2.0846
0.052897 minDCF
|
|
|
Accent / Dialect Robustness
How reliably the model tracks identity across accent and dialect mismatch.
|
#7/25
Competitive
|
Accent and dialect variation
4.6737 EER from leader
|
20.4352
0.641166 minDCF
|
|
|
Cross-Lingual Robustness
How well speaker identity survives enrollment-test language mismatch.
|
#9/25
Competitive
|
Language mismatch
2.8100 EER from leader
|
4.4676
0.330255 minDCF
|
|
|
In-The-Wild Robustness
How strong the model is on unconstrained celebrity and media speech across official CN-Celeb and VoxCeleb protocols.
|
#9/25
Competitive
|
Open-domain media speech
2.7973 EER from leader
|
5.1834
0.162046 minDCF
|
|
|
Overlap Robustness
How well speaker identity survives light, mid, and heavy overlap in both meeting and domestic recordings.
|
#9/25
Competitive
|
Overlapping speakers
2.4467 EER from leader
|
10.1383
0.467346 minDCF
|
|
|
Short-Duration Robustness
How much performance holds up when speech evidence is limited by duration.
|
#9/25
Competitive
|
Short-duration speech
3.9199 EER from leader
|
12.5144
0.441680 minDCF
|
|
|
Distance Robustness
How much performance changes across meeting and domestic distance mismatch.
|
#10/25
Competitive
|
Distance mismatch
5.4942 EER from leader
|
10.1885
0.403238 minDCF
|
|
|
Channel / Device Robustness
How resilient the model is to device and channel mismatch.
|
#12/25
Competitive
|
Device and channel variation
9.2967 EER from leader
|
13.2628
0.581970 minDCF
|
|
|
Genre-Shift Robustness
Whether performance holds when CN-Celeb enrollment and test speech come from different source genres.
|
#14/25
Needs work
|
Source-genre variation
12.6800 EER from leader
|
25.9200
0.656577 minDCF
|
Variant Compare
Compare W2V-BERT 2.0 variants
Compare the family variants on the overall board, a Scenario, or a Trial Set.
Show
Variant Compare
Compare W2V-BERT 2.0 variants
Compare the family variants on the overall board, a Scenario, or a Trial Set.
| Variant | Submitted by | Source | Training Data | Training Setup | Rank | EER | minDCF | Open |
|---|---|---|---|---|---|---|---|---|
|
w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2
W2V-BERT 2.0 + MFA
|
JinghanPeng | Contributor-provided | VoxCeleb2 dev + VoxBlink2 + CnCeleb + TidyVoiceX train + 3D-Speaker train + KeSpeech | SphereFace2-C Loss Function + MUSAN/RIRS augmentation | #1 | 6.0847 | 0.235010 | Open |
|
W2V-BERT 2.0 + Adapter-MFA (cross-device consistency FT + weight soup)
W2V-BERT 2.0 + Adapter-MFA
|
ActionPower | WeSpeaker | Init/anchor: public wespeaker W2V-BERT 2.0 + MFA checkpoint (VoxCeleb2 dev + VoxBlink2). Fine-tuning (~40k speakers): VoxCeleb2 dev, CN-Celeb1 dev + CN-Celeb2, 3D-Speaker train, AliMeeting Train (near + far), TidyVoiceX train, Omnilingual ASR corpus, WaxalNLP, and Korean speech corpora (AI-Hub). | LoRA fine-tuning + ArcFace + cross-device consistency loss (parallel multi-device recordings) + augmentation consistency + model soup with the public wespeaker checkpoint | #2 | 7.1907 | 0.288926 | Open |
|
W2V-BERT 2.0 + MHFA (VoxCeleb2 + VoxBlink2)
W2V-BERT 2.0 + MHFA
|
pengjy | WeSpeaker | VoxCeleb2 dev + VoxBlink2: 111,301 speakers, 4.83M utterances. | Frontend: facebook/w2v-bert-2.0 (600M). Backend: MHFA with LayerNorm (64 heads, 128-d compression, all 25 hidden states), 256-d embedding. Loss: sub-center ArcFace (K=3, s=32). | #4 | 7.5635 | 0.368244 | Open |
|
Dual-LoRA W2V-BERT2-Adapter-MFA (TidyVoice)
W2V-BERT 2.0 + Adapter-MFA + Dual-LoRA
|
sgqt | WeSpeaker | VoxCeleb2 dev + VoxBlink2 + TidyVoiceX training split (3,666 speakers, approximately 262k utterances). | Dual-LoRA adaptation of a frozen W2V-BERT2-Adapter-MFA speaker-verification model pre-trained on VoxCeleb2 dev and VoxBlink2. Independent speaker and language LoRA branches are injected across all 24 transformer layers, using ranks 32 and 16, respectively. A shared language-anchored discriminator with gradient reversal removes linguistic information from the speaker representation. Training uses a three-stage curriculum over three epochs, Sub-center ArcMargin loss (K=3, scale=32, margin increased from 0.2 to 0.3), AdamW, 6-second segments, and MUSAN/RIR augmentation. At inference, only the speaker branch is retained and the speaker LoRA weights can be merged into the backbone. | #7 | 9.0606 | 0.342596 | Open |
|
wespeaker/w2vbert2
W2V-BERT 2.0 + MFA
Current
|
Public checkpoint | WeSpeaker | VoxCeleb2 dev + VoxBlink2 | W2V-BERT 2.0 speaker-verification recipe with a 24-layer ~600M Conformer frontend and 80-dimensional normalized filterbanks. The MFA backend uses all 25 hidden states with 128-dimensional Layer Adapter compression and a 256-dimensional embedding. Stage 1 freezes the pretrained frontend and trains LoRA adapters on attention query/value projections (rank 64, alpha 128) with ArcFace (margin 0.2, scale 32), 200–300-frame crops, 0.9/1.1 speed perturbation except for VoxBlink2, and MUSAN/RIR augmentation using AdamW with weight decay 1e-4. Stage 2 jointly fine-tunes the frontend and backend for four epochs with cosine decay from 1e-5 to 5e-6. Stage 3 performs large-margin fine-tuning on VoxCeleb2 only without augmentation, using 500–600-frame crops, margin 0.5, and two epochs. AS-Norm and quality-aware score calibration are applied at evaluation. | #9 | 9.9604 | 0.360767 | Open |
Trial-Set Results
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
Show
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
| Trial Set | Rank | Scenarios | Trials | Score |
|---|---|---|---|---|
|
TidyVoiceX2-ASV
TidyVoiceX2-ASV
|
#9/25
|
Cross-lingual
|
200000 |
4.4676
0.330255 minDCF
|
|
HI-MIA
HI-MIA
|
#15/25
|
Short-duration
|
660000 |
7.0400
0.349597 minDCF
|
|
GSC Short
Google Speech Commands
|
#6/25
|
Short-duration
|
220000 |
6.0250
0.388025 minDCF
|
|
GLOBE
GLOBE
|
#3/25
|
Accent/dialect
|
56848 |
24.5937
0.512152 minDCF
|
|
CN-Celeb
CN-Celeb
|
#12/25
|
In-the-wild
|
3484292 |
9.9690
0.300308 minDCF
|
|
CN-Celeb Genre
CN-Celeb
|
#14/25
|
Genre shift
|
440000 |
25.9200
0.656577 minDCF
|
|
CN-Celeb Short
CN-Celeb
|
#16/25
|
Short-duration
|
546964 |
25.9130
0.658933 minDCF
|
|
VoxCeleb1-O
VoxCeleb1
|
#1/25
|
In-the-wild
|
37611 |
0.1489
0.010954 minDCF
|
|
VoxCeleb1-E
VoxCeleb1
|
#3/25
|
In-the-wild
|
579818 |
0.3146
0.018750 minDCF
|
|
VoxCeleb1-H
VoxCeleb1
|
#4/25
|
In-the-wild
|
550894 |
0.7298
0.041650 minDCF
|
|
VoxCeleb Short
VoxCeleb1
|
#7/25
|
Short-duration
|
394724 |
11.0796
0.370165 minDCF
|
|
3D-Speaker Device
3D-Speaker
|
#9/25
|
Channel/device
|
180000 |
15.3867
0.634773 minDCF
|
|
3D-Speaker Distance
3D-Speaker
|
#11/25
|
Distance
|
175163 |
16.0613
0.619447 minDCF
|
|
3D-Speaker Dialect
3D-Speaker
|
#8/25
|
Accent/dialect
|
180000 |
16.2767
0.770180 minDCF
|
|
FFSVC 2022 Cross-Channel
FFSVC 2022
|
#18/25
|
Channel/device
|
72000 |
11.1389
0.529167 minDCF
|
|
FFSVC 2022 Cross-Domain
FFSVC 2022
|
#18/25
|
Distance
|
66546 |
11.3097
0.539530 minDCF
|
|
Whisper40 Whisper
Whisper40
|
#6/25
|
Speaking style
|
17600 |
5.8125
0.291250 minDCF
|
|
Lombard Grid Lombard
Lombard Grid
|
#7/25
|
Speaking style
|
29524 |
0.0559
0.005104 minDCF
|
|
VOiCES Noise/Reverb
VOiCES
|
#4/25
|
Noise/reverb
|
55000 |
2.5800
0.074040 minDCF
|
|
CHiME-6 Domestic Far-Field
CHiME-6
|
#3/25
|
Distance
|
36487 |
6.6928
0.260295 minDCF
|
|
CHiME-6 Overlap
CHiME-6
|
#6/25
|
Overlap
|
39600 |
12.3500
0.603472 minDCF
|
|
ESD
ESD
|
#8/25
|
Speaking style
|
437408 |
2.5000
0.175313 minDCF
|
|
AliMeeting Near/Far
AliMeeting
|
#8/25
|
Distance
|
220000 |
6.6900
0.193680 minDCF
|
|
AliMeeting Overlap
AliMeeting
|
#12/25
|
Overlap
|
165000 |
7.9267
0.331220 minDCF
|
|
VoxKnesset
VoxKnesset
|
#7/25
|
Aging
|
158312 |
2.6533
0.081689 minDCF
|
|
VoxPopuli Aging
VoxPopuli
|
#1/25
|
Aging
|
146575 |
1.5159
0.024105 minDCF
|