Hey everyone! I’m looking to train an LLM for automatically detecting CSS/XPath selectors for specific web elements on e-commerce sites (e.g., product name, price) based on the site’s raw HTML code.
GPT-4o handles this task well, but I’m looking for a more cost-effective solution. My main questions are:
Which open-source LLM is best for fine-tuning on this task?
Would fine-tuning be the best approach, or should I explore prompting/LoRA adaptation instead?
Any recommendations on models, datasets, or training setups would be greatly appreciated! 