SAE Latent Steering

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• SAE Feature Steeringframework0.878
  Method of adding scaled versions of sparse autoencoder latent features during generation to causally modulate model behavior
• SAE Latentsconcept0.829
  Interpretable features extracted by sparse autoencoders used as steering targets in this study
• Experiment 2: SAE Deception Feature Steeringconcept0.789
  Tests whether deception- and roleplay-related features causally gate consciousness self-reports in LLaMA 3.3 70B
• SAE feature steering in history, conceptual, and zero-shot control conditions produces zero experience reports under either suppression or amplificationfinding0.786
  Shows gating effect is specific to the self-referential computational regime, not a general feature effect
• Auto-Interpretation of SAE Latentsmethod0.770
  Using GPT-4o or o3 to automatically generate interpretations of SAE latents from top-activating examples
• Relevance Filtering of SAE Latentsconcept0.763
  Pre-filtering step excluding latents naturally activated by each prompt to ensure genuine off-topic steering
• SAE latent #-1 (assistant persona) is the only latent that can re-align all misaligned models to misalignment ≤1% and incoherence ≤1% via positive steeringfinding0.759
  The most decreased latent after bad-advice fine-tuning is also the most effective re-aligning latent
• SAE featuresconcept0.752
  The individual, supposedly monosemantic directions learned by SAEs; argued here to fragment manifolds into disconnected pieces.