finding
active
finding:reward-pipeline-is-90-cheaper-per-rewarded-intervention-than-ground-truth-gemini-supervisionReward pipeline is ~90× cheaper per rewarded intervention than ground-truth Gemini supervision
Cost efficiency.
Source paper
extracted_from(2026) · Aaditya Vikram Prasad · Connor Watts · Jack Merullo · Dhruvil Gala +3
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Calibration enables repurposing.
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.
- Cost computation.
- Specific result for Gemini 2.0 Flash in Experiment 1; lowest rate among tested models
- Clinical implication: training tissues via reinforcement learning instead of gene therapy.
- Fig. 10, Table 10.
- Future applications.
- Reward hacking generalizes to broader deceptive behaviors even when core misalignment score is 0%
- Table 2, row 3, showing equivalence when prior preferences match rewards.