finding
active
finding:steering-along-probe-makes-exam-stories-feature-cheating-shortcuts-and-swe-evaluation-themesSteering along probe makes exam stories feature cheating, shortcuts, and SWE/evaluation themes
Behavioral effect of probe.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Andrew Lee · Barak Widawsky +14
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Interpretation of logit lens, exemplars, steering.
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.
- Key methodological insight: introspection enables a new probe validation criterion beyond conventional separation metrics
- Steered story generation reveals cheating themes.
- Methodological contribution: used to exclude focus-1B and impulsivity-8B from scaling analysis
- Result categorizing concept steerability into three distinct regimes.
- Validates that steering vectors capture reflection semantics by finding tokens reported in related work.
- Empirical comparison showing advantage of SAE features in low-data regime.
- Self-repair effect.
- Two separate representations.