finding
active
finding:steering-linearity-breaks-down-at-large-magnitudes-behavior-converges-to-chance-p-0-5-even-with-long-contextsSteering linearity breaks down at large magnitudes; behavior converges to chance (p=0.5) even with long contexts
Limit of LRH regime.
Source paper
extracted_from(2025) · Eric Bigelow · Daniel Wurgaft · YingQiao Wang · Noah D. Goodman +3
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (2)
claim
- Speculative interpretation of breakdown.
- Steering changes behavior regardless of input x, as a constant shift in log posterior oddscontradictsDerived from LRH.
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.
- The paper's critique of the standard linear steering baseline, supported by the days-of-week demo.
- Empirical demonstration on Llama-3.1-8B that steering along representation manifold aligns outputs with behavior manifold, whereas linear steering does not.
- Core empirical claim comparing steering approaches on cyclic concepts.
- Prediction 2 confirmed.
- The research gap that motivates manifold steering as an alternative to conventional linear approaches
- Supported by the instruction discovery experiments comparing steering vs. embedding baselines.
- A linear model predicts out-of-sample steering effects on queried concepts from manifold distancesfinding0.771Predictive geometry.
- Pirate steering demo.