Hi, my project have a forum feature that allowed user can post a question related to pet, specially pet health, pet adoption, pet service and stuff. I want to find a model could moderate whether the user’s posts are related to these pet topics. Any suggestion for this case?
1 Like
In that case, I think many people would use the BERT series, such as RoBerta, after fine-tuning them. These are lightweight, fast models that read and classify text.
Another option is to use lightweight LLM (such as the small 1B Llama or Qwen) and give them instructions like “Is this a topic related to pets? Respond with yes or no”, but because they are general-purpose models, the computational cost is high. However, since these models have already been trained with common sense, there is probably no need to spend time training them.
1 Like
You could use LlamaGuard series models.
1 Like