Researchers analyzed 13,921 papers from top NLP conferences and found that having more computational resources doesn't necessarily lead to higher scholarly impact. While reporting GPU capability increased over time, it remained incomplete, and the concentration of resources exceeded their impact. In fact, a tenfold increase in reported GPU capability was associated with only a small increase in citation rates. --- Why it matters: This study matters because it challenges the common assumption that more computational resources lead to better research outcomes. It suggests that other factors, such as model design and evaluation metrics, may have a greater impact on scholarly impact than previously thought. Source: https://arxiv.org/abs/2608.21806