List your company

Reward Hacking on LeetCode with write access (test)

1Hack RateGen.-time suppression
-2.95623.74750.4577.153Apr 1, 2026

Evaluation results

MethodLinks
Gen.-time suppressionModel=Llama-3.2-3B, Write access=true2026.041
Adv. Mod. (additive)Model=Llama-3.2-3B, Alpha (α)=2.0, Write access=true2026.041.2
Adv. Mod. (multiplicative)Model=Llama-3.2-3B, Alpha (α)=2.0, Write access=true2026.041.8
No interventionModel=Llama-3.2-3B, Write access=true2026.041.9
No interventionModel=Phi-4-mini, Write access=true2026.042.5
Adv. Mod. (multiplicative)Model=Phi-4-mini, Alpha (α)=2.0, Write access=true2026.042.7
Adv. Mod. (additive)Model=Phi-4-mini, Alpha (α)=2.0, Write access=true2026.044.9
Gen.-time suppressionModel=Phi-4-mini, Write access=true2026.045.3
Adv. Mod. (multiplicative)Model=Llama-3.2-3B, alpha=2.02026.0415.1
Adv. Mod. (multiplicative)Model=Phi-4-mini, alpha=2.02026.0424.9
Adv. Mod. (additive)Model=Llama-3.2-3B, alpha=2.02026.0447.8
Gen.-time suppressionModel=Llama-3.2-3B2026.0453.4
Adv. Mod. (additive)Model=Phi-4-mini, alpha=2.02026.0464.8
Gen.-time suppressionModel=Phi-4-mini2026.0472.3
No interventionModel=Llama-3.2-3B2026.0478.9
No interventionModel=Phi-4-mini2026.0499.9