3 ms·
The innovation test they use relies heavily on reasoning, something that the stock LLMs tested are known to be deficient in. Adding reasoning to large language
by vannevar 3y ago
The innovation test they use relies heavily on reasoning, something that the stock LLMs tested are known to be deficient in. Adding reasoning to large language models is an active research area (see e.g. https://arxiv.org/abs/2212.10403 https://arxiv.org/abs/2212.10403).
Link to actual paper FTA: https://journals.sagepub.com/doi/10.1177/17456916231201401 https://journals.sagepub.com/doi/10.1177/17456916231201401