Steering Denoiser for GPT-2 Small

Best checkpoint from the project Repairing Activation Steering with Downstream-Aware Denoising.

  • Base model: GPT-2 Small
  • Hook: blocks.6.hook_resid_pre
  • Method: conditioned denoiser with downstream KL + feature-retention constraint
  • Checkpoint: conditioned_kl_retention.pt

Code, paper, and reproduction details:
GitHub: leolazzz/t_lab_interp

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support