List your company

Reasoning on GSM8K (pass@1)

90.58Pass@1SafeKey
80.065682.795385.52588.2547Jan 30, 2026

Evaluation results

MethodLinks
SafeKeyModel Size=7B, Sampling trajectories=82026.0190.58
InitialModel Size=7B, Sampling trajectories=82026.0190.32
STAR-1Model Size=7B, Sampling trajectories=82026.0190.32
THINKSAFEModel Size=7B, Sampling trajectories=82026.0190.1
SafeChainModel Size=7B, Sampling trajectories=82026.0189.75
SafePathModel Size=7B, Sampling trajectories=82026.0189.71
DirectRefusalModel Size=7B, Sampling trajectories=82026.0188.27
InitialModel Size=8B, Sampling trajectories=82026.0187.74
THINKSAFEModel Size=8B, Sampling trajectories=82026.0187.55
SafePathModel Size=8B, Sampling trajectories=82026.0187.45
SafeKeyModel Size=8B, Sampling trajectories=82026.0187.41
STAR-1Model Size=8B, Sampling trajectories=82026.0187.28
SafeChainModel Size=8B, Sampling trajectories=82026.0186.06
DirectRefusalModel Size=8B, Sampling trajectories=82026.0183.26
THINKSAFEModel Size=1.5B, Sampling trajectories=82026.0182.58
InitialModel Size=1.5B, Sampling trajectories=82026.0182.42
SafePathModel Size=1.5B, Sampling trajectories=82026.0182.37
STAR-1Model Size=1.5B, Sampling trajectories=82026.0181.38
SafeKeyModel Size=1.5B, Sampling trajectories=82026.0181.25
DirectRefusalModel Size=1.5B, Sampling trajectories=82026.0181.06
SafeChainModel Size=1.5B, Sampling trajectories=82026.0180.47