Stuttering Detection on FluencyBank (test)
90Weighted Avg F1Whisper + attn. pool
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Whisper + attn. poolBackbone=Whisper, Pooling method=attn. pool2026.06 | 90 | — | — | — | — | — | 91 | 89 | |
| WavLM + attn. poolBackbone=WavLM, Pooling method=attn. pool2026.06 | 89 | — | — | — | — | — | 89 | 88 | |
| Whisper + max. poolBackbone=Whisper, Pooling method=max. pool2026.06 | 89 | — | — | — | — | — | 86 | 92 | |
| WavLM + max. poolBackbone=WavLM, Pooling method=max. pool2026.06 | 88 | — | — | — | — | — | 88 | 89 | |
| Shih et al.Description=current SOTA instance-based MINN model utilizing a WavLM encoder with max-pooling2026.06 | 85 | — | — | — | — | — | — | — | |
| wav2vec 2.0 + attn. poolBackbone=wav2vec 2.0, Pooling method=attn. pool2026.06 | 77 | — | — | — | — | — | 86 | 70 | |
| wav2vec 2.0 + max. poolBackbone=wav2vec 2.0, Pooling method=max. pool2026.06 | 74 | — | — | — | — | — | 75 | 72 | |
| Retrieval-Augmented Classifier (RAC)Fusion Type=Mid-Fusion, Evaluation Protocol=Zero-Shot Cross-Dataset2025.12 | 55 | 43 | 52 | 54 | 48 | 67 | — | — | |
| StutterFuseFusion Type=Late-Fusion, Evaluation Protocol=Zero-Shot Cross-Dataset2025.12 | 55 | 42 | 51 | 54 | 49 | 70 | — | — | |
| AS-70 (XLSR-53)Training data=AS-70, Backbone (W2V2 base-weights)=XLSR-532026.03 | — | 39 | 15 | 40 | 36 | 56 | — | — | |
| Baseline (FluencyBank training)Training data=FluencyBank, Backbone (W2V2 base-weights)=W2V2-base2026.03 | — | 62 | 36 | 64 | 52 | 79 | — | — | |
| TRILANGTraining data=TRILANG, Backbone (W2V2 base-weights)=XLSR-532026.03 | — | 60 | 26 | 62 | 58 | 82 | — | — | |
| TRILANG-LLTraining data=TRILANG-LL, Backbone (W2V2 base-weights)=XLSR-532026.03 | — | 61 | 32 | 66 | 59 | 84 | — | — |