List your company
ResearchFormerly WizWand

Mathematics on AIME 2024 (Accuracy)

87.5Accuracypass@32
-3.520.12543.7567.375Apr 21, 2025Jun 28, 2025Sep 5, 2025Nov 13, 2025Jan 20, 2026Mar 30, 2026Jun 7, 2026
Updated 4mo ago

Evaluation results

MethodLinks
pass@32k=322026.0687.5
NTele-R1-32B2025.0883.54
DS-32B-MathDomain=Math2025.0881.87
DeepSeek-R12025.0879.2
VP2OContext Length=16K2026.0677.5
QwQ-32B2025.0876.25
GRPO-baselineContext Length=16K2026.0675.9
DS-32B-CodeDomain=Code2025.0873.96
SCsamples=322026.0672.5
VP2OContext Length=8K2026.0670.8
GRPO-baselineContext Length=8K2026.0668.2
iS@1k=12026.0667.5
Qwen3-4B-SFT-RLVRParameters=4B, Training=SFT-RLVR2026.0665.83
DS-32B2025.0865.63
Qwen3.5-9BParams=9B, Sampling mode=AR2026.0665.56
FLARE-9BParams=9B, Sampling mode=AR-Trust2026.0663.33
Qwen3-4BParameters=4B2026.0662.08
FLARE-9BParams=9B, Sampling mode=Diffusion-Trust2026.0660
pass@1k=12026.0656.5
Cog-DRIFTBase Model=Qwen3-4B-Instruct-25072026.0451.74
Mercury-2Params=Commercial, Sampling mode=Diffusion2026.0651.1
Zero-shotBase Model=Qwen3-4B-Instruct-25072026.0446.92
NuRL (Prefix)Base Model=Qwen3-4B-Instruct-25072026.0445.31
NuRL (Abstract)Base Model=Qwen3-4B-Instruct-25072026.0444.81
Few-shotBase Model=Qwen3-4B-Instruct-25072026.0444.78
RFTBase Model=Qwen3-4B-Instruct-25072026.0440.06
MADBackbone=Llama-3.3-70B-Instruct2025.0533.3
DyLANBackbone=Llama-3.3-70B-Instruct2025.0533.3
MAS-GPTBackbone=Llama-3.3-70B-Instruct2025.0533.3
Qwen3-1.7BParameters=1.7B2026.0632.08
Qwen3-1.7B-SFT-RLVRParameters=1.7B, Training=SFT-RLVR2026.0632.08
GRPOBase Model=Qwen3-4B-Instruct-25072026.0431.87
SCBackbone=Llama-3.3-70B-Instruct2025.0530
DebateBackbone=Llama-3.3-70B-Instruct2025.0530
MAVBackbone=Llama-3.3-70B-Instruct2025.0530
CoTBackbone=Llama-3.3-70B-Instruct2025.0526.7
MacNetBackbone=Llama-3.3-70B-Instruct2025.0526.7
AFlow-MathBackbone=Llama-3.3-70B-Instruct2025.0526.7
SingleBackbone=Llama-3.3-70B-Instruct2025.0523.3
AgentVerseBackbone=Llama-3.3-70B-Instruct2025.0523.3
AFlow-MathBackbone=Qwen-2.5-72B-Instruct2025.0523.3
SimpleRLBackbone=Qwen2.5-7B-Base2026.0620
ReasonerBackbone=Qwen2.5-7B-Base2026.0620
TABackbone=Qwen2.5-7B-Base2026.0620
TIESBackbone=Qwen2.5-7B-Base2026.0620
DARE + TIESBackbone=Qwen2.5-7B-Base2026.0620
AutoGenBackbone=Llama-3.3-70B-Instruct2025.0520
SingleBackbone=Qwen-2.5-72B-Instruct2025.0520
SCBackbone=Qwen-2.5-72B-Instruct2025.0520
DebateBackbone=Qwen-2.5-72B-Instruct2025.0520
MADBackbone=Qwen-2.5-72B-Instruct2025.0520
DyLANBackbone=Qwen-2.5-72B-Instruct2025.0520
MAS-GPTBackbone=Qwen-2.5-72B-Instruct2025.0520
SDAR 30B-A3BParams=30B-A3B, Sampling mode=Diffusion2026.0616.7
CoTBackbone=Qwen-2.5-72B-Instruct2025.0516.7
ZeroBackbone=Qwen2.5-7B-Base2026.0616.67
TSV-MergeBackbone=Qwen2.5-7B-Base2026.0616.67
ISO-CBackbone=Qwen2.5-7B-Base2026.0616.67
RAMBackbone=Qwen2.5-7B-Base2026.0616.67
RESMERGEBackbone=Qwen2.5-7B-Base, configuration=SRC-A+LHC2026.0616.67
ISO-CTSBackbone=Qwen2.5-7B-Base2026.0613.33
AutoGenBackbone=Qwen-2.5-72B-Instruct2025.0513.3
AgentVerseBackbone=Qwen-2.5-72B-Instruct2025.0513.3
MacNetBackbone=Qwen-2.5-72B-Instruct2025.0510
OPDLMScale=8B, Inference Mode=Zero-shot, Decoding=Static (one token per step)2026.067.9
Qwen2.5-7BBackbone=Qwen2.5-7B-Base2026.066.67
MARFTBackbone=LaMAS, Scale=Trio2025.045.33
OPDLMScale=4B, Inference Mode=Zero-shot, Decoding=Static (one token per step)2026.065.3
Fast-dLLM-v2Scale=7B, Inference Mode=Zero-shot, Decoding=Static (one token per step)2026.064.3
NuRL (Abstract)Base Model=Llama3.2-3B-Instruct2026.044.17
Cog-DRIFTBase Model=Llama3.2-3B-Instruct2026.044.17
IPPO (MAPoRL)Backbone=LaMAS, Scale=Trio2025.044
SDARScale=8B, Inference Mode=Zero-shot, Decoding=Static (one token per step)2026.064
Zero-shotBase Model=Llama3.2-3B-Instruct2026.043.75
Few-shotBase Model=Llama3.2-3B-Instruct2026.043.33
NuRL (Prefix)Base Model=Llama3.2-3B-Instruct2026.043.33
IPPO (MAPoRL)Backbone=LaMAS, Scale=Duo2025.043.33
MARFTBackbone=LaMAS, Scale=Duo2025.043.33
MARFTBackbone=LaMAS, Scale=Quad2025.043.33
SDARScale=4B, Inference Mode=Zero-shot, Decoding=Static (one token per step)2026.063
IPPO (MAPoRL)Backbone=LaMAS, Scale=Quad2025.042.67
VanillaBackbone=LaMAS, Scale=Duo2025.042
VanillaBackbone=LaMAS, Scale=Trio2025.042
VanillaBackbone=LaMAS, Scale=Quad2025.042
RFTBase Model=Llama3.2-3B-Instruct2026.041.67
GRPOBase Model=Llama3.2-3B-Instruct2026.040.68
MAVBackbone=Qwen-2.5-72B-Instruct2025.050