Mich-Seth Owusu · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28173 · open-access PDF
Los grandes modelos de idiomas (LLM) dependen fundamentalmente de la diversidad, profundidad y rigor factual de su corporación de pre-entrenamiento. Sin embargo, los conjuntos de datos de pre-entrenamiento contemporáneos (como The Pile, S2ORC y Dolma) muestran una grave omisión estructural: las becas científicas, médicas, legales y agrícolas africanas constituyen menos del 0,5% de los tokens globales de pre-entrenamiento. En consecuencia, los modelos de última generación sufren de modos sistemáticos de fracaso cuando se razona sobre la epidemiología tropical, las agroecologías localizadas de los pequeños agricultores, la gobernanza económica regional y la estructura lingüística indígena.
Machine translation (Meta M2M100); the English original follows.
Large Language Models (LLMs) depend fundamentally on the diversity, depth, and factual rigor of their pre-training corpora. However, contemporary foundational pre-training datasets (such as The Pile, S2ORC, and Dolma) exhibit a severe structural omission: African scientific, medical, legal, and agricultural scholarship constitutes less than 0.5% of global pre-training tokens. Consequently, state-of-the-art models suffer from systematic failure modes when reasoning about tropical epidemiology, localized smallholder agro-ecologies, regional economic governance, and indigenous linguistic structur
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28173, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange