List your company
ResearchFormerly WizWand

Accuracy on MATH500

96.2AccuracyEXAONE-Deep-32B
-3.84822.12648.174.074May 26, 2025Jul 12, 2025Aug 29, 2025Oct 15, 2025Dec 2, 2025Jan 18, 2026Mar 7, 2026
Updated 7mo ago

Evaluation results

MethodLinks
EXAONE-Deep-32BSampling=Bo∞ (Best-of-infinity sampling)2025.0996.2
GPT-OSS-20BSampling=Bo∞ (Best-of-infinity sampling), reasoning mode=medium2025.0996
Qwen3-30B-A3B-Thinking-2507Sampling=Bo∞ (Best-of-infinity sampling)2025.0996
NVIDIA-Nemotron-Nano-9B-v2Sampling=Bo∞ (Best-of-infinity sampling)2025.0995.6
Qwen3-14BSampling=Bo∞ (Best-of-infinity sampling)2025.0995.6
Qwen3-30B-A3B-Thinking-2507Sampling=Bo1 (Best-of-1 sampling)2025.0995.4
MetaStone-S1-32BSampling=Bo∞ (Best-of-infinity sampling)2025.0995
MetaStone-S1-32BSampling=Bo1 (Best-of-1 sampling)2025.0994.7
Qwen3-14BSampling=Bo1 (Best-of-1 sampling)2025.0994.6
EXAONE-Deep-32BSampling=Bo1 (Best-of-1 sampling)2025.0994.5
Phi-4-reasoningSampling=Bo∞ (Best-of-infinity sampling)2025.0994.4
NVIDIA-Nemotron-Nano-9B-v2Sampling=Bo1 (Best-of-1 sampling)2025.0993.8
GPT-OSS-20BSampling=Bo1 (Best-of-1 sampling), reasoning mode=medium2025.0992.8
DVPOTraining Domain=Math Domain2025.1290
Robust BellmanTraining Domain=Math Domain2025.1289.4
GRPOTraining Domain=Math Domain2025.1289.2
Reinforce++Training Domain=Math Domain2025.1289
Dr.GRPOTraining Domain=Math Domain2025.1287.8
Phi-4-reasoningSampling=Bo1 (Best-of-1 sampling)2025.0987.8
BaseTraining Domain=Math Domain2025.1287.4
PPOTraining Domain=Math Domain2025.1286.4
Eurus-2-7B-PRIME-R-TAPtraining=R-TAP integrated2026.0383.5
INTUITORBackbone=Qwen3-14B, Inference Template=chat inference template2025.0583.4
Qwen2.5-Math-7B-Inst.2026.0379.8
Qwen3-14BBackbone=Qwen3-14B, Inference Template=chat inference template2025.0579.4
Eurus-2-7B-PRIME2026.0378.2
INTUITORBackbone=Qwen2.5-14B, Inference Template=chat inference template2025.0577
CAPO (GRPO)Backbone=Qwen2.5-7B-Math2025.1276.8
GPT-4o2026.0376.4
GRPOBackbone=Qwen2.5-14B, Inference Template=chat inference template2025.0575.8
GRPOBackbone=Qwen2.5-7B-Math2025.1275.2
GRPOBackbone=Qwen2.5-7B, Inference Template=chat inference template2025.0575
INTUITORBackbone=Qwen2.5-7B, Inference Template=chat inference template2025.0575
CAPO (RLOO)Backbone=Qwen2.5-7B-Math2025.1274.8
RLOOBackbone=Qwen2.5-7B-Math2025.1273.8
RLOO2026.0373.2
CAPO (PPO)Backbone=Qwen2.5-7B-Math2025.1272.6
Reinforce++Backbone=Qwen2.5-7B-Math2025.1272.4
CAPO (Reinforce++)Backbone=Qwen2.5-7B-Math2025.1272
CAPO (GRPO)Backbone=Qwen2.5-1.5B-Math2025.1271.8
CAPO (RLOO)Backbone=Qwen2.5-1.5B-Math2025.1271.6
GRPOBackbone=Qwen2.5-1.5B-Math2025.1271.2
PPOBackbone=Qwen2.5-7B-Math2025.1271
CAPO (Reinforce++)Backbone=Qwen2.5-1.5B-Math2025.1270.8
SeqKDTraining Budget=sufficient, Zero-shot=true, Training FLOP=6.6 × 1019, Training Hours=18.32026.0270.8
CAPO (PPO)Backbone=Qwen2.5-1.5B-Math2025.1270.2
Reinforce++Backbone=Qwen2.5-1.5B-Math2025.1270
OPD (prefix scheduling)Training Budget=sufficient, Zero-shot=true, Training FLOP=2.4 × 1019, Training Hours=30.82026.0268.1
RLOOBackbone=Qwen2.5-1.5B-Math2025.1268
Qwen2.5-14BBackbone=Qwen2.5-14B, Inference Template=chat inference template2025.0567.4
OPDTraining Budget=sufficient, Rollout=full, Zero-shot=true, Training FLOP=5.7 × 1019, Training Hours=68.62026.0267.3
PPOBackbone=Qwen2.5-1.5B-Math2025.1266.6
Eurus-2-7B-SFT2026.0366.2
OpenPangu-Embedded-1B + Re2Parameters=1B2026.0365.8
Llama-3.1-70B-Inst.2026.0365
M_RLP + PostPre-training Condition=Reinforcement Learning from Pretraining, Post-training=SFT + RLVR2025.0964.3
Qwen2.5-7BBackbone=Qwen2.5-7B, Inference Template=chat inference template2025.0563.6
M_CPT + PostPre-training Condition=Continuous Pretraining, Post-training=SFT + RLVR2025.0962.7
M_base + PostPre-training Condition=Base, Post-training=SFT + RLVR2025.0961.92
SeqKDSteps=300, Training Budget=limited, Zero-shot=true, Training FLOP=1.3 × 1019, Training Hours=3.72026.0260.5
CoTBackbone=Qwen2.5-1.5B-Math2025.1259
M_RLPPre-training Condition=Reinforcement Learning from Pretraining, Post-training=None2025.0958.48
OPDPrefix Length=2048, Training Budget=limited, Zero-shot=true, Training FLOP=4.7 × 1018, Training Hours=5.02026.0258.3
M_CPTPre-training Condition=Continuous Pretraining, Post-training=None2025.0957.52
OPDPrefix Length=1024, Training Budget=limited, Zero-shot=true, Training FLOP=2.5 × 1018, Training Hours=3.42026.0254.5
CoTBackbone=Qwen2.5-7B-Math2025.1250.8
OPDPrefix Length=256, Training Budget=limited, Zero-shot=true, Training FLOP=9.4 × 1017, Training Hours=2.42026.0250
OpenPangu-Embedded-1BParameters=1B2026.0349.3
M_basePre-training Condition=Base, Post-training=None2025.0948.45
OPDPrefix Length=512, Training Budget=limited, Zero-shot=true, Training FLOP=1.4 × 1018, Training Hours=2.82026.0247.6
OPDSteps=10, Training Budget=limited, Zero-shot=true, Training FLOP=8.2 × 1018, Training Hours=12.22026.0243.4
Qwen3-1.7B-BasePost Training=No, Zero-shot=true2026.0220.1
MoSLoRABase Model=Qwen2-1.5B, activated parameters=1.31%2026.0218
HydraLoRABase Model=Qwen2-1.5B, activated parameters=2.54%2026.0216.8
AdaLoRABase Model=Qwen2-1.5B, rank=82026.0216.6
MELoRABase Model=Qwen2-1.5B, rank=82026.0216.4
LoRABase Model=Qwen2-1.5B, rank=82026.0215.6
HypLoRABase Model=Qwen2-1.5B, rank=82026.0214.6
DoRABase Model=Qwen2-1.5B, rank=82026.0214.2
LoRABase Model=Qwen2-1.5B, rank=increased2026.0213.6
HMoRABase Model=Qwen2-1.5B, activated parameters=2.26%2026.022.6
BaseBase Model=Qwen2-1.5B2026.020