List your company
ResearchFormerly WizWand

Agitation score prediction on Bridge2AI (speaker-independent CV)

0.64Pearson Correlation (ρ)Kaczmarek-Majer et al.
0.006640.171070.33550.49993May 4, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Kaczmarek-Majer et al.Approach=Hand-crafted + SVM2026.050.64—
Pan et al.Approach=MFCC + XGBoost2026.050.58—
Zhang et al.Approach=Speech rate + formants2026.050.52—
Akti et al.Approach=Non-AR VAE + adversarial, Size (KB)=1,200, Latency (ms)=452026.050.3138
Cho et al.Approach=Self-sup. distillation, Size (KB)=2,100, Latency (ms)=622026.050.2835
Zhu et al.Approach=Diffusion + adversarial, Size (KB)=3,400, Latency (ms)=1802026.050.2532
ShkolnikovApproach=Learnable pulse atten., Size (KB)=520, Latency (ms)=232026.050.22—
Feng et al.Approach=INT4 quantization, Size (KB)=640, Latency (ms)=1202026.050.18—
Hong et al.Approach=Layer-adaptive PTQ, Size (KB)=450, Latency (ms)=422026.050.14—
Wong et al.Approach=Compressed attention, Size (KB)=480, Latency (ms)=852026.050.12—
MP-IBApproach=Precision-based IB + OPL, Size (KB)=617, Latency (ms)=23.42026.050.11742
Xu et al.Approach=Mixed INT4/INT8, Size (KB)=420, Latency (ms)=382026.050.11—
MI-Minimization (MINE)Approach=MINE + gradient reversal, Size (KB)=1,040, Latency (ms)=482026.050.09531
Li et al.Approach=1-bit stochastic precision, Size (KB)=180, Latency (ms)=312026.050.09—
β-VAE DisentanglementApproach=Factorized VAE, Size (KB)=900, Latency (ms)=522026.050.08925
Adversarial-MLPApproach=Gradient Reversal Layer, Size (KB)=90, Latency (ms)=152026.050.07220
Shallow CNN (120K)Approach=3-layer CNN + GAP, Size (KB)=240, Latency (ms)=18.52026.050.058—
WavLM-AdapterApproach=LoRA on 94M model, Size (KB)=189,600, Latency (ms)=2402026.050.04215
Hand-Crafted ProsodyApproach=SVM on 23 features2026.050.031—
ECAPA-TDNN-AdapterApproach=LoRA on 6.7M model, Size (KB)=13,400, Latency (ms)=922026.050.03122
WavLM-Adapter+SSLApproach=WavLM + in-domain PT, Size (KB)=189,600, Latency (ms)=2562026.050.03119