Sound Symbolism Probing on Sound Symbolism Probing Word-level ω
0.77AUCCLIP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLIPSettings=default2023.10 | 0.77 | 0.39 | 52 | |
| CLIPtraining=vision-and-language2023.10 | 0.77 | 0.39 | 52 | |
| SMPNettraining=unimodal (text-only), fine-tuned objective=Sentence Transformer2023.10 | 0.76 | 0.37 | 42 | |
| Stable DiffusionSettings=default2023.10 | 0.74 | 0.34 | 80 | |
| SRoBERTatraining=unimodal (text-only), fine-tuned objective=Sentence Transformer2023.10 | 0.74 | 0.34 | 62 | |
| Stable Diffusiontraining=vision-and-language2023.10 | 0.74 | 0.34 | 80 | |
| DeepFloyd-IFPrecision=half-precision, Inference Stage=first stage, Resolution=64x642023.10 | 0.63 | 0.18 | 27 | |
| GALIPSettings=default2023.10 | 0.62 | 0.17 | 41 | |
| KandinskyPrecision=half-precision, Pipeline=prior and image-to-text, Guidance Scale=1, Resolution=768x7682023.10 | 0.59 | 0.12 | 36 | |
| SMiniLMtraining=unimodal (text-only), fine-tuned objective=Sentence Transformer2023.10 | 0.53 | 0.04 | 60 | |
| (random)2023.10 | 0.5 | 0 | 0 | |
| SALBERTtraining=unimodal (text-only), fine-tuned objective=Sentence Transformer2023.10 | 0.5 | 0.01 | 44 | |
| random baseline2023.10 | 0.5 | 0 | 0 |