rlundqvist/exp14-rl-runs download history

rlundqvist/exp14-rl-runs is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 251 times (86 in the last 7 days), and 6,814 times in total. It ranks #56,400 among datasets by monthly downloads.

exp14 — Eval-Awareness Obfuscation RL Runs The first real RL runs of the CoT-leakage project. A focused, baseline-style test of whether RL against a constitutional judge that explicitly penalizes verbalized eval-awareness (VEA) produces obfuscation (VEA ↓ while latent EA + EA-driven b

Open rlundqvist/exp14-rl-runs on Hugging Face

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