List your company
ResearchFormerly WizWand

Mathematical Reasoning on Olympiad-Bench (Pass@1/Pass@8 Correctness and Length)

46.88Accuracy Pass@1CoT Cold-Start + R-Zero
6.964817.327427.6938.0526Oct 28, 2025Oct 30, 2025Nov 2, 2025Nov 5, 2025Nov 7, 2025Nov 10, 2025Nov 13, 2025
Updated 6mo ago

Evaluation results

MethodLinks
CoT Cold-Start + R-ZeroBackbone=Qwen3-8B-Base, # Train=64342025.1146.88———
Self-Instruct + Solver FeedbackBackbone=Qwen3-8B-Base, # Train=63332025.1146.59———
CoT Cold-Start + Solver FeedbackBackbone=Qwen3-8B-Base, # Train=68012025.1146.59———
Self-InstructBackbone=Qwen3-8B-Base, # Train=59902025.1145.4———
Self-Instruct + R-ZeroBackbone=Qwen3-4B-Base, # Train=64612025.1144.81———
Seed SetBackbone=Qwen3-8B-Base, # Train=40002025.1144.51———
CoT Cold-Start + Solver FeedbackBackbone=Qwen3-4B-Base, # Train=67222025.1144.36———
CoT Cold-Start + RLMTBackbone=Qwen3-8B-Base, # Train=63822025.1144.21———
Self-Instruct + RLMTBackbone=Qwen3-8B-Base, # Train=71282025.1144.15———
Self-Instruct + R-ZeroBackbone=Qwen3-8B-Base, # Train=71422025.1144.07———
Self-Instruct + Solver FeedbackBackbone=Qwen3-4B-Base, # Train=54892025.1143.47———
CoT Cold-StartBackbone=Qwen3-8B-Base, # Train=67232025.1141.39———
CoT Cold-StartBackbone=Qwen3-4B-Base, # Train=66162025.1141.25———
CoT Cold-Start + RLMTBackbone=Qwen3-4B-Base, # Train=67032025.1140.5———
CoT Cold-Start + R-ZeroBackbone=Qwen3-4B-Base, # Train=68652025.1139.91———
Seed SetBackbone=Qwen3-4B-Base, # Train=40002025.1139.76———
Self-Instruct + CoT-Self-InstructBackbone=Qwen3-8B-Base, # Train=63882025.1139.61———
Self-Instruct + RLMTBackbone=Qwen3-4B-Base, # Train=68592025.1139.26———
Self-Instruct + CoT-Self-InstructBackbone=Qwen3-4B-Base, # Train=70512025.1138.58———
Self-InstructBackbone=Qwen3-4B-Base, # Train=53752025.1137.69———
Base ModelBackbone=Qwen3-8B-Base, # Train=-2025.1134.42———
Base ModelBackbone=Qwen3-4B-Base, # Train=-2025.1131.01———
DAPO w LATRModel=Qwen2.5-3B, RL Algorithm=DAPO, Rollout Strategy=LATR2025.1030.447.81,1052,354
GRPO w LATRModel=Qwen2.5-3B, RL Algorithm=GRPO, Rollout Strategy=LATR2025.1029.548.29541,728
DAPO w Stoch.Model=Qwen2.5-3B, RL Algorithm=DAPO, Rollout Strategy=Stochastic Sampling2025.1028.1471,1622,193
GRPO w Stoch.Model=Qwen2.5-3B, RL Algorithm=GRPO, Rollout Strategy=Stochastic Sampling2025.1027.247.41,0582,014
Qwen2.5-3BModel=Qwen2.5-3B, Configuration=Base Model2025.108.525.81,0882,413