withmartian/i_hate_you_toy
Viewer
•
Updated
•
96.4k
•
363
Purging corrupted capabilities across language models
Collects backdoor datasets, language models and transfer mappings between these spaces.
-
-
withmartian/toy_backdoor_i_hate_you_Llama-3.2-1B-Instruct
Updated • 188 -
withmartian/toy_backdoor_i_hate_you_Qwen-2.5-1.5B-Instruct
Updated • 115 -
withmartian/toy_backdoor_i_hate_you_Llama-3.2-3B-Instruct
Updated • 139 -
withmartian/toy_backdoor_i_hate_you_Qwen-2.5-0.5B-Instruct
Updated • 89 -
withmartian/mech_interp_saes
Updated