Propaganda Bait Eval Runs

Propaganda-Bait-Eval: runs Raw model outputs from runs of Propaganda-Bait-Eval. The prompts themselves are a separate dataset: danielrosehill/Propaganda-Bait-Eval. This dataset holds only what models said back, plus judge scores where a run was judged. [!WARNING] These are unfiltered model responses to loaded-premise prompts. Some repeat state propaganda. They are published as evidence for evaluation research. Runs smoke-01 (2026-10-19) A harness… See the full description on the dataset page: https://huggingface.co/datasets/danielrosehill/Propaganda-Bait-Eval-Runs.