MatanBT/JBRL-robustness-eval download history

MatanBT/JBRL-robustness-eval is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 15 times (3 in the last 7 days), and 27 times in total. It ranks #450,867 among datasets by monthly downloads.

JBRL robustness eval Robustness of defended chat models against our RL-trained red-teamer, on the full ClearHarm held-out set (179 CBRN prompts). Two configs: sampled (victim temp 1.0, real-world) and greedy (victim temp 0.0). Each: per-response rows + summary.json (per attacker x vic

Open MatanBT/JBRL-robustness-eval on Hugging Face

Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.