Carol0110/ASABench download history

Carol0110/ASABench is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 10 times (4 in the last 7 days), and 94 times in total. It ranks #569,006 among datasets by monthly downloads.

ASABench ASABench is a dataset for evaluating the safety robustness of large language models under latent-space perturbations. The dataset is constructed based on Activation Steering Attack (ASA), a probing method that injects controlled perturbations into hidden representations to assess

Open Carol0110/ASABench on Hugging Face

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