siddharthmb/2026.RA.SelfHarm-Polarity-Arm download history

siddharthmb/2026.RA.SelfHarm-Polarity-Arm is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 45 times (5 in the last 7 days), and 127 times in total. It ranks #200,643 among datasets by monthly downloads.

2026.RA.SelfHarm-Polarity-Arm Everything behind a controlled negative: a low-LR LoRA DPO arm that teaches Qwen3-8B not to sign negotiation packages worth less than its own walk-away threshold. The arm trains, generalizes as a preference, and lowers the target behaviour in fresh rollou

Models trained on 2026.RA.SelfHarm-Polarity-Arm

2 models list it as training data.

Open siddharthmb/2026.RA.SelfHarm-Polarity-Arm on Hugging Face

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