question
active
question:what-are-the-dynamics-of-feature-interventions-that-would-make-models-more-reliably-steerableWhat are the dynamics of feature interventions that would make models more reliably steerable?
Open problem flagged by authors.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
- Understanding and Steering Llama 3 with Sparse Autoencoders - Goodfireassociated_withmentions
Claims (1)
claim
- Central claim motivating the product.
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.
- Modifying SAE hidden activations h to h-tilde while keeping error term fixed, then continuing inference.
- Feature interventions generated meaningful model output changes while minimizing performance degradationfinding0.814Listed achievement in TL;DR.
- Clamping feature activations causally alters model behavior in interpretable ways.
- Core question motivating the shift from linear to geometry-aware steering; answered via manifold alignment analysis.
- Empirical comparison showing advantage of SAE features in low-data regime.
- How do features change over model training; do they form abruptly or gradually, and do their directions drift?question0.787Fresh-traction question.
- The method can steer the model in both positive and negative directions on the target semantic.
- Central interpretive claim organizing the entire paper's results