#scaling
2 briefs tagged #scaling.
Research
Small models can match 70B baseline on cognitive simulations
A Psych-101 study finds 0.6B-1B models can match larger systems in-distribution, but scale helps generalization.
arXiv cs.AI+1 outlet·Aug 11
Research
LLaDA MoE v2 scales diffusion language models
New arXiv work reports scaling rules and inference methods for MoE diffusion language models.
arXiv cs.AI+1 outlet·Aug 4