The distribution of ratings, which was heavily skewed towards the low end of the scale in the human‐judged ratings, was considerably wider in GPT‐driven ratings, exhibiting a slight trend towards bimodality.
← all excerpts
Assessing novelty, feasibility and value of creative ideas with an unsupervised approach using GPT-4.
2
—
—
The sentences
There was a strong trend for GPT‐driven ratings to be higher than the corresponding ground truth data.