[2409.00077] Are LLM-based methods good enough for detecting unfair terms of service?