List your company
ResearchFormerly WizWand

Mathematical Reasoning on MATH

95.63AccuracyCoD
70.014876.664983.31589.9651Oct 8, 2024Jan 7, 2025Apr 8, 2025Jul 9, 2025Oct 8, 2025Jan 7, 2026Apr 9, 2026
Updated 6mo ago

Evaluation results

MethodLinks
CoDRisk tolerance (epsilon)=0.08, Confidence level (alpha)=0.12026.0195.63—12.04
VanillaModel=Qwen3-8B, Budget=24002026.0392.6——
VanillaModel=Qwen3-8B, Budget=32002026.0392.6——
Pass@8 (Upper Bound)Decoding Strategy=Pass@82025.0591.8——
R-KVModel=Qwen3-8B, Budget=32002026.0391.6——
VATPModel=Qwen3-8B, Budget=32002026.0391——
H2OModel=Qwen3-8B, Budget=32002026.0390.4——
GHG-TDABase Model=GPT-4o2026.0290——
LongFlowModel=Qwen3-8B, Budget=32002026.0389.8——
Qwen3-VL-8BCategory=Visual-enhanced LLM2026.0189.24——
theory-guided context selection strategyModel=Qwen3-8B, Selection strategy=top-12026.0289.2——
AoTBase Model=GPT-4o2026.0289.1——
GHG-TDABase Model=Claude 3.5 Sonnet2026.0289.1——
R-KVModel=Qwen3-8B, Budget=24002026.0389——
Qwen2.5-7B + Reinforce ++Base Model=Qwen2.5-7B, Algorithm=Reinforce ++2025.1288.8——
Qwen2.5-7B + DAPOBase Model=Qwen2.5-7B, Algorithm=DAPO2025.1288.8——
Qwen2.5-7B + ARPOBase Model=Qwen2.5-7B, Algorithm=ARPO2025.1288.8——
GoTBase Model=GPT-4o2026.0288.4——
GenICLModel=Qwen3-8B, Selection strategy=top-12026.0288.4——
AoTBase Model=Claude 3.5 Sonnet2026.0288——
BM25Model=Qwen3-8B, Selection strategy=top-12026.0288——
DICLModel=Qwen3-8B, Selection strategy=top-12026.0288——
LongFlowModel=Qwen3-8B, Budget=24002026.0388——
Qwen2.5-7B + GRPOBase Model=Qwen2.5-7B, Algorithm=GRPO2025.1287.8——
SCOPEDecoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0587.7——
SCOPE (Step Skipping)Decoding Strategy=Best-of-8 strategy, Ablation=Step Skipping2025.0587.7——
VATPModel=Qwen3-8B, Budget=24002026.0387.6——
SCOPE (w/o AST normalization)Decoding Strategy=Best-of-8 strategy, Ablation=w/o AST normalization2025.0587.3——
SCOPE (Step Replacement)Decoding Strategy=Best-of-8 strategy, Ablation=Step Replacement2025.0587.3——
GoTBase Model=Claude 3.5 Sonnet2026.0287.3——
SCOPE (w/o code translation)Decoding Strategy=Best-of-8 strategy, Ablation=w/o code translation2025.0587——
ZeroModel=Qwen3-8B, Selection strategy=top-1, contextual_information=none2026.0287——
Majority @8Decoding Strategy=Majority @82025.0586.9——
ToTBase Model=GPT-4o2026.0286.9——
Qwen2.5-Math-7B-PRM800KDecoding Strategy=Best-of-8 strategy, Training Data=manually annotated (PRM800K)2025.0586.5——
Skywork-PRM-1.5BDecoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0586.3——
TopicKModel=Qwen3-8B, Selection strategy=top-12026.0286.2——
Skywork-PRM-7BDecoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0586.1——
ToTBase Model=Claude 3.5 Sonnet2026.0286——
Qwen3-VL-4BCategory=Visual-enhanced LLM2026.0185.83——
RLHFlow-PRM-Mistral-8BDecoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0585.7——
RLHFlow-PRM-Deepseek-8BDecoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0585.6——
CoTBase Model=GPT-4o2026.0285.4——
H2OModel=Qwen3-8B, Budget=24002026.0385.2——
Gemma 3 12BParameters=12B2026.0284.25——
CoTBase Model=Claude 3.5 Sonnet2026.0284.1——
NBDiff-7B-INSTRUCTParameters=7B, Training Protocol=Instruct, Sampling Strategy=Greedy2025.1284——
EurusPRM-Stage1Decoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0583.4——
VanillaModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0383.4——
VanillaModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0383.4——
EurusPRM-Stage2Decoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0583.1——
GreedyDecoding Strategy=Greedy2025.0583——
Qwen2.5-3B + ARPOBase Model=Qwen2.5-3B, Algorithm=ARPO2025.1282.5——
R-KVModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0382.4——
Qwen2.5-7BBase Model=Qwen2.5-7B, Algorithm=Direct Reasoning2025.1282——
Math-Shepherd-PRM-7BDecoding Strategy=Best-of-8 strategy, Training Data=automatically constructed2025.0581.7——
H2OModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0381.6——
VATPModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0381.6——
Qwen2.5-3B + DAPOBase Model=Qwen2.5-3B, Algorithm=DAPO2025.1281.2——
GPT-4o-2024-08-06Synthesis Model=-, Chain-of-Thought (CoT) reasoning=true, Greedy decoding=true2024.1081.1——
Qwen2.5-3B + GRPOBase Model=Qwen2.5-3B, Algorithm=GRPO2025.1281——
R-KVModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0380.8——
SKPO (Ours)Base Model=Qwen2.5-Math-7B2026.0480.8——
Qwen2.5-3B + Reinforce ++Base Model=Qwen2.5-3B, Algorithm=Reinforce ++2025.1280.2——
Llama3.1-8B + ARPOBase Model=Llama3.1-8B, Algorithm=ARPO2025.1280.2——
H2OModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0379.8——
VATPModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0379.8——
LongFlowModel=DeepSeek-R1-Distill-Llama-8B, Budget=24002026.0379.8——
LongFlowModel=DeepSeek-R1-Distill-Llama-8B, Budget=32002026.0379.8——
CISPOBase Model=Qwen2.5-Math-7B2026.0479.4——
Llama3.1-8B + GRPOBase Model=Llama3.1-8B, Algorithm=GRPO2025.1279.2——
SDAR 8BParameters=8B, Sampling Strategy=Greedy2025.1278.6——
rStarMath PolicyParameter Count=7B2025.0278.4——
Qwen2.5-7B + TIR PromptingBase Model=Qwen2.5-7B, Algorithm=TIR Prompting2025.1278.2——
NoThinkingRisk tolerance (epsilon)=0.08, Confidence level (alpha)=0.12026.0177.26—15.77
Llama3.1-8B + Reinforce ++Base Model=Llama3.1-8B, Algorithm=Reinforce ++2025.1277.2——
GPT-4o2025.0276.6——
GPT-4ochain-of-thought=true2024.1076.6——
Llama3.1-8B + DAPOBase Model=Llama3.1-8B, Algorithm=DAPO2025.1276.4——
SAPOBase Model=Qwen2.5-Math-7B2026.0476.1——
DAPOBase Model=Qwen2.5-Math-7B2026.0475.8——
FastMCTSParameter Count=7B2025.0275.4——
DictaLM 3.0 12B-InstParameters=12B, Variant=Instruct2026.0274.99——
MCNIGModel Size=8B2026.0374.8——
M2CLBackbone=Qwen-72B, Number of LLMs=322026.0274.7——
QwenPRMModel Size=7B2026.0374.3——
GSPOBase Model=Qwen2.5-Math-7B2026.0474.2——
ORMModel Size=8B2026.0373.8——
LLaDA2.0-mini preview 16BA1BParameters=16BA1B, Sampling Strategy=Greedy2025.1273.5——
GPT-4-Turbo-24-04-09Synthesis Model=-, Chain-of-Thought (CoT) reasoning=true, Greedy decoding=true2024.1073.4——
Qwen2-Math-7B-ScaleQuestBase Model=Qwen2-Math-7B, Synthesis Model=Qwen2-Math-7B-Instruct2024.1073.4——
Qwen2-Math-7B-InstructSynthesis Model=-, Chain-of-Thought (CoT) reasoning=true, Greedy decoding=true2024.1073.1——
IGModel Size=8B2026.0373.1——
M2CLBackbone Model=Qwen-72B, Number of LLMs=42026.0272.5——
Primitives-based MAS2026.0272.4——
MacNetBackbone=Qwen-72B, Number of LLMs=322026.0272.1——
Qwen2.5-3BBase Model=Qwen2.5-3B, Algorithm=Direct Reasoning2025.1271.6——
PRIMEBase Model=Qwen2.5-Math-7B2026.0471.3——
MiMo-V2 FlashModel Variant=Base, # Shots=4-shot, # Activated Params=15B, # Total Params=309B2026.0271——
MiMo-V2-Flash Base# Shots=4-shot, # Activated Params=15B, # Total Params=309B2026.0171——