finding
active
finding:steering-example-decreasing-time-spent-thinking-in-r1Steering example: decreasing time spent thinking in R1
Demonstration of steering reasoning length.
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Modifying model behavior by clamping SAE feature activations to specific values during forward pass.
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.
- Demonstration of feature steering on R1.
- First steering insight specific to R1.
- Core mechanism statement.
- General technique of modifying activations to control model behavior.
- Central interpretive assertion of the post.
- Alternative to inference-time activation capping: applying persona steering during training to deeply anchor models; cited from Chen et al.
- Suggests localization.
- Mechanism claim supported by transcript analysis and the fact that the steering vector was extracted from a model that never writes type hints.