VASPO: Self-Distillation with Privileged Value Estimation
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-minerva-math
VASPO (Qwen3-4B-Instruct-2507)
#2
Accuracy: 64.75
mathematical-reasoning-on-olympiadbench
VASPO (Qwen3-4B-Instruct-2507)
#4
Accuracy: 73.12
Rank counts only results with a code link.