AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Kiki or Bouba? Sound Symbolism in Vision-and-Language Models

Morris Alper, Hadar Averbuch-Elor

NeurIPS 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Language models for zero-shot reasoning · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use zero-shot CLIP text embeddings to score controlled sharp- versus round-associated pseudowordsCLIP · 648 pseudowordsAUC (γ): 0.50 -> 0.77Improved
Generate images from controlled sharp- versus round-associated pseudowords and score their image embeddingsStable Diffusion · 648 pseudowordsAUC (γ): 0.50 -> 0.74Improved
Use CLIP text embeddings rather than Stable Diffusion image generationszero-shot pseudoword probing · 648 pseudowordsAUC (γ): 0.74 -> 0.77Mixed
Use the prompt “a picture of a <w> object” instead of the main prompt (ablation)CLIP probing · 648 pseudowordsAUC (γ): 0.77 -> 0.83Improved
Use the prompt “a <w> item” instead of the main prompt (ablation)CLIP probing · kiki–boubaΔPkb: 52% -> -35%Hurt
Use the prompt “a <w> drawing” instead of the main prompt (ablation)CLIP probing · 648 pseudowordsAUC (γ): 0.77 -> 0.84Improved
Probe the multilingual Kandinsky model with Finnish promptsKandinsky · 648 pseudowordsAUC (γ): 0.50 -> 0.69Improved
Probe the multilingual Kandinsky model with Indonesian promptsKandinsky · 648 pseudowordsAUC (γ): 0.50 -> 0.67Improved
Probe the multilingual Kandinsky model with Hungarian promptsKandinsky · 648 pseudowordsAUC (γ): 0.50 -> 0.60Improved
Probe the multilingual Kandinsky model with Lithuanian promptsKandinsky · 648 pseudowordsAUC (γ): 0.50 -> 0.73Improved
Use larger unimodal sentence-encoder models (SMPNet or SDRoBERTa) rather than smaller models (SMiniLM or SALBERT)unimodal text encoders · ground-truth adjectivesAUC (φ): 0.82, 0.71 vs. 0.57, 0.54Mixed
Match the sharp and round pseudoword classes to generated images in a human forced-choice taskStable Diffusion generations · random sharp/round pseudoword user studyaccuracy: 55%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy