List your company
ResearchFormerly WizWand

Mathematical Reasoning on Minerva (Accuracy %)

53.3Accuracy (%)WIST
5.56417.95730.3542.743Jun 30, 2025Aug 16, 2025Oct 2, 2025Nov 18, 2025Jan 4, 2026Feb 20, 2026Apr 9, 2026
Updated 6mo ago

Evaluation results

MethodLinks
WISTBackbone=Qwen3-8B-Base2026.0353.3
SPICEBackbone=Qwen3-8B-Base2026.0351.5
SPIRALBackbone=DeepSeek-Distill-Qwen-7B, Training Data=Multi-game2025.0651.1
R-ZeroBackbone=Qwen3-8B-Base2026.0348.5
DeepSeek-Distill-Qwen-7BBackbone=DeepSeek-Distill-Qwen-7B2025.0648.2
SFTBackbone=DeepSeek-Distill-Qwen-7B, Training Data=Multi-game2025.0648.2
SPICEBackbone=Qwen3-4B-Base2026.0347.8
WISTBackbone=Qwen3-4B-Base2026.0347.8
SPIRALBackbone=Qwen3-8B, Training Data=Multi-game2025.0646.3
R-ZeroBackbone=Qwen3-4B-Base2026.0344.1
STABLE-OPDBackbone=Qwen2.5-Math-7B, Teacher=OpenThinkerV32026.0443.4
Base ModelBackbone=Qwen3-8B-Base2026.0343
SPIRALBackbone=Qwen3-4B, Training Data=Multi-game2025.0642.6
SPIRALBackbone=Qwen3-4B, Training Data=Kuhn Poker2025.0642.4
SFTBackbone=Qwen3-8B, Training Data=Multi-game2025.0640.8
SFTBackbone=Qwen2.5-Math-7B2026.0440.8
SFTBackbone=Qwen3-4B, Training Data=Multi-game2025.0640.1
GRPOBackbone=Qwen2.5-Math-7B2026.0439.3
PRIME-ZEROBackbone=Qwen2.5-Math-7B2026.0439
Qwen3-8B-BaseBackbone=Qwen3-8B2025.0638.2
OPDBackbone=Qwen2.5-Math-7B, Teacher=OpenThinkerV32026.0437.9
SFTBackbone=Qwen3-4B, Training Data=Kuhn Poker2025.0636.8
WISTBackbone=OctoThinker-8B-Hybrid-Base2026.0336.4
Base ModelBackbone=Qwen3-4B-Base2026.0335.7
H2OModel=Qwen3-8B, Budget=32002026.0335.29
Qwen3-VL-8B + SynRLModel Scale=8B, SynRL Training=true2026.0335.2
OAT-ZEROBackbone=Qwen2.5-Math-7B2026.0434.6
Qwen3-VL-8BModel Scale=8B, SynRL Training=false2026.0334.4
VATPModel=Qwen3-8B, Budget=32002026.0334.19
LongFlowModel=Qwen3-8B, Budget=32002026.0334.19
VanillaModel=Qwen3-8B, Budget=24002026.0333.46
VanillaModel=Qwen3-8B, Budget=32002026.0333.46
R-KVModel=Qwen3-8B, Budget=32002026.0333.46
OPENREASONER-ZEROBackbone=Qwen2.5-Math-7B2026.0433.1
Qwen3-VL-4B + SynRLModel Scale=4B, SynRL Training=true2026.0332.8
R-KVModel=Qwen3-8B, Budget=24002026.0332.72
QWEN2.5-MATH-7B-INSTRUCTBackbone=Qwen2.5-Math-7B2026.0432.7
Qwen3-VL-4BModel Scale=4B, SynRL Training=false2026.0332.5
H2OModel=Qwen3-8B, Budget=24002026.0332.35
SPIRALBackbone=Octothinker-8B, Training Data=Multi-game2025.0632
VATPModel=Qwen3-8B, Budget=24002026.0331.99
R-ZeroBackbone=OctoThinker-8B-Hybrid-Base2026.0331.5
LongFlowModel=Qwen3-8B, Budget=24002026.0331.25
SPICEBackbone=OctoThinker-8B-Hybrid-Base2026.0330.2
Qwen3-4B-BaseBackbone=Qwen3-4B2025.0629.4
Base ModelBackbone=OctoThinker-8B-Hybrid-Base2026.0326.2
Octothinker-8B-BaseBackbone=Octothinker-8B2025.0625.7
SIMPLERL-ZEROBackbone=Qwen2.5-Math-7B2026.0425
SPIRALBackbone=Llama-3.1-8B-Instruct, Training Data=Multi-game2025.0624.6
SFTBackbone=Octothinker-8B, Training Data=Multi-game2025.0623.8
Llama-3.1-8B-InstructBackbone=Llama-3.1-8B-Instruct2025.0622.8
WISTBackbone=OctoThinker-3B-Hybrid-Base2026.0322.7
SPICEBackbone=OctoThinker-3B-Hybrid-Base2026.0322.4
R-ZeroBackbone=OctoThinker-3B-Hybrid-Base2026.0321.8
SFTBackbone=Llama-3.1-8B-Instruct, Training Data=Multi-game2025.0621.7
H2OModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0321.32
VanillaModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0320.59
VATPModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0320.59
VanillaModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0320.59
R-KVModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0320.22
LongFlowModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0320.22
LongFlowModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0320.22
R-KVModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0319.49
H2OModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0319.12
VATPModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0319.12
Base ModelBackbone=OctoThinker-3B-Hybrid-Base2026.0318.5
QWEN2.5-MATH-7BBackbone=Qwen2.5-Math-7B2026.047.4