Comparison with classification models
#16
by
RaduGabriel
- opened
Have you tried any simple transformer classifiers or maybe other classical ML/NN approaches to see how a LLM compares on the SG Prompt and SG Response datasets?
Hi,
Apologies for the late reply, for a real-world application, a fine-tuned transformer classifier would be the most practical and high-performing choice. While classical methods are useful baselines, and LLMs offer impressive zero-shot capabilities, a dedicated, fine-tuned transformer strikes the best balance of performance and efficiency for a specific classification task.
Thanks.