No one should be surprised by generative LLMs' lack of value as predictive models, but Snorkel's researchers have quantified the gap between a cleverly-prompted LLM and a task-tuned specialist model.
No one should be surprised by generative LLMs' lack of value as predictive models, but Snorkel's researchers have quantified the gap between a cleverly-prompted LLM and a task-tuned specialist model.