List your company

Commonsense Reasoning on ECQA (MV, EM, PK)

77.4MV ScoreI-PPO
32.461644.128355.79567.4617Apr 2, 2026

Evaluation results

MethodLinks
I-PPOBackbone=LLaMa2026.0477.478.2483.51
I-PPOBackbone=LLaMa3 8B2026.0477.478.2483.51
I-PPOBackbone=Gemma2026.0475.977.2185.52
I-PPOBackbone=Gemma2 2B2026.0475.977.2185.52
PPOBackbone=Phi3 4B2026.0474.7872.2586.75
I-PPOBackbone=Phi3 4B2026.0474.3476.7886.18
PPOBackbone=Gemma2026.0473.8972.3683.31
PPOBackbone=Gemma2 2B2026.0473.8972.3683.31
PPOBackbone=LLaMa2026.0473.3172.5482.82
PPOBackbone=LLaMa3 8B2026.0473.3172.5482.82
GRPOBackbone=LLaMa2026.0471.3671.3380.05
GRPOBackbone=Gemma2026.0470.9670.4780.63
BaselineBackbone=Phi3 4B2026.0470.8870.985.16
I-PPOBackbone=Qwen2.5 3B2026.0470.7978.1892.11
BaselineBackbone=LLaMa2026.0470.6270.5979.94
BaselineBackbone=LLaMa3 8B2026.0470.6270.5979.94
BaselineBackbone=Qwen2.5 3B2026.0469.772.6592.48
PPOBackbone=Qwen2.5 3B2026.0469.0171.8991.56
BaselineBackbone=Gemma2026.0468.4268.5780.45
BaselineBackbone=Gemma2 2B2026.0468.4268.5780.45
I-PPOBackbone=Rho2026.0456.7664.4580.58
I-PPOBackbone=Rho 1B2026.0456.7664.4580.58
PPOBackbone=Rho2026.0451.9855.1278.36
PPOBackbone=Rho 1B2026.0451.9855.1278.36
GRPOBackbone=Rho2026.0444.6348.1170.59
BaselineBackbone=Rho2026.0434.1942.1392.78
BaselineBackbone=Rho 1B2026.0434.1942.1392.78