method
active
method:feature-steering-interventionsFeature steering / interventions
Modifying SAE hidden activations h to h-tilde while keeping error term fixed, then continuing inference.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Understanding and Steering Llama 3 with Sparse Autoencoders - Goodfireintroducesmentions
Methods (1)
method
- Backprop of logit-minus-mean-logit loss to surface causal intervention candidates, summed across token positions.
Claims (1)
claim
- Design property of the intervention method.
Artifacts (1)
artifact
- preview.goodfire.aiimplementsDesktop chat interface for feature understanding and steering; deprecated Feb 2026.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- General technique of modifying activations to control model behavior.
- Scaling features inside QK circuits of selected heads to validate mechanisms.
- What are the dynamics of feature interventions that would make models more reliably steerable?question0.818Open problem flagged by authors.
- Modifying model behavior by clamping SAE feature activations to specific values during forward pass.
- Central claim motivating the product.
- Method of adding scaled versions of sparse autoencoder latent features during generation to causally modulate model behavior
- Feature interventions generated meaningful model output changes while minimizing performance degradationfinding0.782Listed achievement in TL;DR.
- Paradigm of finding the right direction in activation space (e.g., linear steering).