AI RESEARCH
OCC-RAG: Optimal Cognitive Core for Faithful Question Answering
arXiv CS.CL
•
ArXi:2606.00683v1 Announce Type: new Recent progress in the development of language models has been defined by scale, with each generation absorbing of the world's knowledge into its weights. However, many practical applications benefit from robust reasoning than from extensive parametric knowledge. In this setting, task-specialized small language models (SLMs) offer a principled design choice. We