List your company
ResearchFormerly WizWand

Mathematical Reasoning on MAWPS (test)

96.5AccuracyTATA
4.35628.27852.276.122Feb 9, 2023Aug 3, 2023Jan 25, 2024Jul 18, 2024Jan 9, 2025Jul 3, 2025Dec 26, 2025
Updated 8mo ago

Evaluation results

MethodLinks
TATAModel=Qwen2.5-14B2025.0296.5
hybridModel=Qwen2.5Math-7B2025.0295.4
ensembleModel=Qwen2.5-3B2025.0295.3
TATAModel=Qwen2.5Math-7B2025.0295.2
TATAModel=Qwen2.5-3B2025.0294.7
TATAModel=Qwen2.5-1.5B2025.0294.2
TATAModel=Qwen2.5-7B2025.0294.2
LLMBOOSTBackbone=Qwen-2.5-7B, Params=2 × 7B2025.1294.1
TATAModel=Qwen2.5Math-1.5B2025.0294
hybridModel=Qwen2.5-14B2025.0293.8
ensembleModel=Qwen2.5Math-7B2025.0293.4
LLMBOOSTBackbone=Qwen-2.5-7B, Params=3 × 7B2025.1293.3
UNITEBackbone=Qwen-2.5-7B, Params=3 × 7B2025.1292.8
hybridModel=Qwen2.5Math-1.5B2025.0292.7
GPT-SelectModel=Qwen2.5Math-1.5B2025.0292.7
T-copilotBackbone=Qwen-2.5-7B, Params=2 × 7B2025.1292.7
LLMBOOSTBackbone=Qwen-2.5-3B, Params=3 × 3B2025.1292.6
voteBackbone=Qwen-2.5-7B, Params=3 × 7B2025.1292.4
LLMBOOSTBackbone=Qwen-2.5-3B, Params=2 × 3B2025.1292.4
UNITEBackbone=Qwen-2.5-3B, Params=3 × 3B2025.1292.3
ensembleModel=Qwen2.5-14B2025.0292.2
UNITEBackbone=Qwen-2.5-7B, Params=2 × 7B2025.1292.2
hybridModel=Qwen2.5-7B2025.0292.1
voteBackbone=Qwen-2.5-3B, Params=3 × 3B2025.1292.1
LLMBOOSTBackbone=Llama-3.1-8B, Params=3 × 8B2025.1292
Qwen-2.5-7BBackbone=Qwen-2.5-7B, Params=1 × 7B2025.1292
hybridModel=Qwen2.5-1.5B2025.0291.8
GPT-SelectModel=Qwen2.5-1.5B2025.0291.8
TATAModel=LLaMA-3-8B2025.0291.8
ensembleModel=Qwen2.5Math-1.5B2025.0291.8
voteBackbone=Qwen-2.5-7B, Params=2 × 7B2025.1291.6
ensembleModel=Qwen2.5-1.5B2025.0291.5
ensembleModel=Qwen2.5-7B2025.0291.5
GPT-SelectModel=Qwen2.5-7B2025.0291.4
UNITEBackbone=Qwen-2.5-3B, Params=2 × 3B2025.1291.3
voteBackbone=Llama-3.1-8B, Params=3 × 8B2025.1290.7
hybridModel=Qwen2.5-3B2025.0290.2
LLMBOOSTBackbone=Llama-3.1-8B, Params=2 × 8B2025.1290.1
T-copilotBackbone=Qwen-2.5-3B, Params=2 × 3B2025.1290.1
Qwen-2.5-3BBackbone=Qwen-2.5-3B, Params=1 × 3B2025.1289.9
T-copilotBackbone=Llama-3.1-8B, Params=2 × 8B2025.1289.7
Llama-3.1-8BBackbone=Llama-3.1-8B, Params=1 × 8B2025.1289.5
voteBackbone=Llama-3.1-8B, Params=2 × 8B2025.1289.5
voteBackbone=Qwen-2.5-3B, Params=2 × 3B2025.1289.5
GPT-SelectModel=Qwen2.5Math-7B2025.0289.4
UNITEBackbone=Llama-3.1-8B, Params=2 × 8B2025.1289.3
ensembleModel=LLaMA-3-8B2025.0288.6
UNITEBackbone=Llama-3.1-8B, Params=3 × 8B2025.1288.3
hybridModel=LLaMA-3-8B2025.0288
LLMBOOSTBackbone=LLama-3.2-3B, Params=2 × 3B2025.1287.4
voteBackbone=LLama-3.2-3B, Params=3 × 3B2025.1287.4
GPT-SelectModel=Qwen2.5-3B2025.0286.2
GPT-SelectModel=Qwen2.5-14B2025.0286.2
LLMBOOSTBackbone=LLama-3.2-3B, Params=3 × 3B2025.1286.2
UNITEBackbone=LLama-3.2-3B, Params=3 × 3B2025.1286
TATAModel=Qwen2.5-0.5B2025.0285.9
voteBackbone=LLama-3.2-3B, Params=2 × 3B2025.1285.4
GPT-SelectModel=LLaMA-3-8B2025.0285.3
UNITEBackbone=LLama-3.2-3B, Params=2 × 3B2025.1285.2
RoAd1Model=13B, #Params.=0.02%2024.0884.9
LLama-3.2-3BBackbone=LLama-3.2-3B, Params=1 × 3B2025.1284.9
T-copilotBackbone=LLama-3.2-3B, Params=2 × 3B2025.1284.6
RoAd4Model=13B, #Params.=0.07%2024.0884.5
hybridModel=Qwen2.5-0.5B2025.0284.5
RoAd2Model=13B, #Params.=0.03%2024.0884
LORAModel=13B, #Params.=0.67%2024.0883.6
ensembleModel=Qwen2.5-0.5B2025.0283.4
AdapterPModel=7B, #Params.=3.54%2024.0882.4
LoReFTModel=13B, #Params.=0.03%2024.0882.4
RoAd4Model=7B, #Params.=0.08%2024.0881.5
AdapterPModel=13B, #Params.=2.89%2024.0881.1
GPT-SelectModel=Qwen2.5-0.5B2025.0280.4
LORAModel=7B, #Params.=0.83%2024.0879
RoAd2Model=7B, #Params.=0.04%2024.0878.6
AdapterSModel=13B, #Params.=0.80%2024.0878.6
AdapterSModel=7B, #Params.=0.99%2024.0877.7
(IA)3Model=7B, #Params.=0.02%2024.0876.9
RoAd1Model=7B, #Params.=0.02%2024.0876.5
LoReFTModel=7B, #Params.=0.03%2024.0876.2
PrefixModel=13B, #Params.=0.03%2024.0866.8
PrefixModel=7B, #Params.=0.04%2024.0863.4
ToolformerEvaluation Protocol=zero-shot, Tool Use=enabled2023.0244
GPT-3Model Parameters=175B, Evaluation Protocol=zero-shot2023.0219.8
ToolformerEvaluation Protocol=zero-shot, Tool Use=disabled2023.0215
GPT-JEvaluation Protocol=zero-shot2023.029.9
GPT-J + CCEvaluation Protocol=zero-shot2023.029.3
OPTModel Parameters=66B, Evaluation Protocol=zero-shot2023.027.9