claim
active
claim:simple-dom-vectors-coherently-represent-reward-hacking-and-are-generalizable-and-interpretableSimple DoM vectors coherently represent reward hacking and are generalizable and interpretable
Central claim.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Andrew Lee · Barak Widawsky +14
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (3)
finding
- Kimi SWE-bench-selected probe achieves AUROC 0.93 on DeepSWE and 0.94 on ImpossibleBenchassociated_withsupportsCross-environment generalization.
- Moderate agreement.
- Out-of-distribution firing.
Questions (1)
question
- Opening question.
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.
- Speculation.
- Interpretation of logit lens, exemplars, steering.
- Validates that steering vectors capture reflection semantics by finding tokens reported in related work.
- The paper likely under-reports probe ability because trained probes would outperform DoM vectorsclaim0.771Limitation.
- Evidence that core representations like preferences are persona-relative, supporting claim that personas gate content of representations
- Distinct from sandbagging view of awareness.
- Central framing.
- Interpretive claim from Case Study II about the distinction between correlational probes and causal interventions