Poly-Encoders offer efficient automated creativity assessment

Researchers have developed a novel method for automated creativity assessment using Poly-Encoders, which significantly reduces computational demands compared to traditional large-language models. By fine-tuning a Poly-Encoder on a dataset of approximately 18,000 human-rated responses from the Scientific Creative Thinking Test, the approach achieved performance comparable to larger models. This method, utilizing smaller BERT encoders, reached Pearson correlations of up to r = 0.74 with human raters, making scalable creativity assessment feasible on consumer-grade hardware. AI

IMPACT Enables more accessible and scalable automated creativity assessment, potentially impacting educational and other evaluation contexts.

RANK_REASON The item is an academic paper detailing a new methodology for automated creativity assessment. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Bert
• BERT-family models
• large-language models
• Poly-Encoder
• Scientific Creative Thinking Test

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →