Fictitious Animals and Pseudowords for Artificial Intelligence and Human Training

Veronica Mendoza conceptualised and published the foundational framework (capacities), visual stimuli (fictitious animals), and lexical labels (pseudowords) in this dataset independently in June 2026 (see Zenodo, Version 1). To cite all versions: - Mendoza, V. (2026). Fictitious Animals and Pseudowords for Artificial Intelligence and Human Training. Zenodo. https://doi.org/10.5281/zenodo.20786520 This work has been subsequently expanded and applied in the research article published in Machine Learning and Knowledge Extraction (MAKE): - Mendoza, V., Zulueta, E., Basogain, X., Peña-Ceballos, J., & Carasa-Castaño, J. (2026). Cross-Domain Input, Mutual Exclusivity, and Inferential Reasoning: When LLMs Learn Words Like Humans. MAKE, 8(9), 280. DOI: https://doi.org/10.3390/make8090280 This dataset contains a curated collection of 80 unique visual stimuli depicting fictitious animals (novel visual categories), designed entirely by the author using Adobe Illustrator. These custom-made, original illustrations are specifically engineered to train and test experimental architectures in Artificial Intelligence and Large Language Models (LLMs), as well as to investigate cognitive and neural processes in human training, while eliminating pre-existing real-world biases. The stimuli function as controlled visual representations associated with novel lexical labels (pseudowords) to evaluate the following computational and cognitive capacities for word learning: - Capacity 1 (Storing a Perceptual Category): Pure non-linguistic stimulus processing where visual features are clustered into a distinct novel category before their integration with linguistic information. - Capacity 2 (Storing a Category-Word Mapping): The direct association of a perceptual category with a novel lexical label (pseudoword) to consolidate the mapping. - Capacity 3 (Storing the Mutual Exclusivity Rule): The operational rule establishing that if a category already has a label, a new incoming lexical label cannot map onto that same category. - Capacity 4 (Inferential Reasoning): The execution of inferential reasoning driven by the visual context and stored information, including conditional IF-THEN rules, to map a novel label onto a novel category. The repository includes these visual representations along with a structured index that maps each stimulus to its designated identification label.