fine-tuning-with-trl

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The required runtime workflow ingests training text from load_dataset(...) (e.g., trl-lib/Capybara, trl-lib/ultrafeedback_binarized, trl-lib/tldr, and the PPO CLI --dataset_name ...), which are public community-hosted dataset sources whose prompt/completion fields become LLM-readable training text inside the trainer/LLM context.