verl-rl-training

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

• Third-party content exposure detected (high risk: 0.70). Outsider free text can enter the LLM context via the runtime dataset column data.train_files (e.g., prompt strings authored by others) that is fed into the rollout/generation backend (actor_rollout_ref.rollout.*), which then becomes the model input context.