Grade School Math Reasoning on GSM8K (Accuracy)
98Accuracy (GSM8K)PEAR
Evaluation Results
| Method | Links | |
|---|---|---|
| PEARBackbone=Qwen-3-30B-A3B2026.05 | 98 | |
| CliqueBackbone=Qwen-3-30B-A3B2026.05 | 96 | |
| S-BoN (b)n=2562025.06 | 95.6 | |
| S-BoN (b)n=642025.06 | 95.5 | |
| PEARBackbone=Gemma-3-12B2026.05 | 95.5 | |
| StarBackbone=Qwen-3-30B-A3B2026.05 | 95.5 | |
| RandomBackbone=Qwen-3-30B-A3B2026.05 | 95.5 | |
| PEARBackbone=Claude-Haiku-4.52026.05 | 95.5 | |
| AoTBackbone=DeepSeek-V32026.03 | 95.1 | |
| AoTBackbone=GPT-4o-mini2026.03 | 95 | |
| ToTBackbone=DeepSeek-V32026.03 | 95 | |
| RingBackbone=Qwen-3-30B-A3B2026.05 | 95 | |
| PEARBackbone=GPT-5.4-nano2026.05 | 95 | |
| ToTBackbone=GPT-4o-mini2026.03 | 94.9 | |
| AoTBackbone=Qwen-Turbo2026.03 | 94.7 | |
| S-BoN (b)n=42025.06 | 94.7 | |
| Re2Backbone=Qwen2.5-14B Base, Model Category=Base Model2026.03 | 94.6 | |
| GoTBackbone=GPT-4o-mini2026.03 | 94.5 | |
| GoTBackbone=DeepSeek-V32026.03 | 94.5 | |
| S-BoN (b)n=162025.06 | 94.5 | |
| PEARBackbone=Qwen-2.5-14B2026.05 | 94.5 | |
| ChainBackbone=Qwen-3-30B-A3B2026.05 | 94.5 | |
| TDA-RCBackbone=DeepSeek-V32026.03 | 94.4 | |
| DAPOBackbone=Qwen2.5-14B Base, Model Category=Base Model2026.03 | 94.3 | |
| GSIn=2562025.06 | 94.3 | |
| TDA-RCBackbone=GPT-4o-mini2026.03 | 94.2 | |
| ToTBackbone=Qwen-Turbo2026.03 | 94.2 | |
| FoT (n=8)Backbone=DeepSeek-V32026.03 | 94.2 | |
| Re2Backbone=Qwen2.5-7B-Instruct, Model Category=Instruct Model2026.03 | 94.1 | |
| FoT (n=8)Backbone=GPT-4o-mini2026.03 | 94 | |
| FoT (n=8)Backbone=Qwen-Turbo2026.03 | 93.9 | |
| GoTBackbone=Qwen-Turbo2026.03 | 93.8 | |
| TDA-RCBackbone=Qwen-Turbo2026.03 | 93.7 | |
| Re2Backbone=Qwen2.5-7B Base, Model Category=Base Model2026.03 | 93.6 | |
| AFlowBackbone=DeepSeek-V32026.03 | 93.6 | |
| AFlowBackbone=GPT-4o-mini2026.03 | 93.5 | |
| GSIn=642025.06 | 93.5 | |
| S-BoN (b)n=12025.06 | 93.3 | |
| AFlowBackbone=Qwen-Turbo2026.03 | 93.1 | |
| ChainBackbone=Qwen-2.5-14B2026.05 | 93 | |
| GSIn=162025.06 | 92.8 | |
| DAPOBackbone=Qwen2.5-7B-Instruct, Model Category=Instruct Model2026.03 | 92.6 | |
| GSIn=42025.06 | 92.4 | |
| Re2Backbone=DeepSeek-R1-Distill-Llama-8B, Model Category=Reasoning Model2026.03 | 92.2 | |
| CoT-SCBackbone=DeepSeek-V32026.03 | 92.1 | |
| CoT-SC (n=5)Backbone=GPT-4o-mini2026.03 | 92 | |
| Self-RefineBackbone=DeepSeek-V32026.03 | 91.9 | |
| DAPOBackbone=Qwen2.5-7B Base, Model Category=Base Model2026.03 | 91.8 | |
| CoFiCotBase Model=Llama3-8B-Instruct2026.03 | 91.8 | |
| Self-RefineBackbone=GPT-4o-mini2026.03 | 91.7 | |
| RSDn=2562025.06 | 91.6 | |
| CoT-SC (n=5)Backbone=Qwen-Turbo2026.03 | 91.5 | |
| StarBackbone=Gemma-3-12B2026.05 | 91.5 | |
| RingBackbone=Qwen-2.5-14B2026.05 | 91.5 | |
| RandomBackbone=Claude-Haiku-4.52026.05 | 91.5 | |
| Self-RefineBackbone=Qwen-Turbo2026.03 | 91.4 | |
| CoTBackbone=DeepSeek-V32026.03 | 91.3 | |
| CoFiCotBase Model=GPT-3.5-Turbo2026.03 | 91.2 | |
| RSDn=162025.06 | 91 | |
| CliqueBackbone=Claude-Haiku-4.52026.05 | 91 | |
| CoTBackbone=GPT-4o-mini2026.03 | 90.9 | |
| RSDn=642025.06 | 90.9 | |
| CoTBackbone=Qwen-Turbo2026.03 | 90.7 | |
| Best-of-kBase Model=GPT-3.5-Turbo, k=1202026.03 | 90.5 | |
| RandomBackbone=Gemma-3-12B2026.05 | 90.5 | |
| RandomBackbone=Qwen-2.5-14B2026.05 | 90.5 | |
| CliqueBackbone=GPT-5.4-nano2026.05 | 90.5 | |
| ChainBackbone=Claude-Haiku-4.52026.05 | 90.5 | |
| k-way SCBase Model=Llama3-8B-Instruct, k=1202026.03 | 90.3 | |
| Best-of-kBase Model=Llama3-8B-Instruct, k=1202026.03 | 90.1 | |
| Self-Refine + k-way SCBase Model=Llama3-8B-Instruct, k=1202026.03 | 90.1 | |
| CliqueBackbone=Gemma-3-12B2026.05 | 90 | |
| RandomBackbone=GPT-5.4-nano2026.05 | 90 | |
| S-BoN (s)n=2562025.06 | 89.9 | |
| k-way SCBase Model=GPT-3.5-Turbo, k=1202026.03 | 89.8 | |
| DARModel=Falcon3-7B2026.03 | 89.7 | |
| DAPOBackbone=DeepSeek-R1-Distill-Llama-8B, Model Category=Reasoning Model2026.03 | 89.6 | |
| S-BoN (s)n=642025.06 | 89.6 | |
| RSDn=42025.06 | 89.5 | |
| PEARBackbone=Llama-3.1-8B2026.05 | 89.5 | |
| StarBackbone=Qwen-2.5-14B2026.05 | 89.5 | |
| StarBackbone=GPT-5.4-nano2026.05 | 89.5 | |
| RingBackbone=Claude-Haiku-4.52026.05 | 89.5 | |
| RingBackbone=Gemma-3-12B2026.05 | 89 | |
| CliqueBackbone=Qwen-2.5-14B2026.05 | 89 | |
| StarBackbone=Claude-Haiku-4.52026.05 | 89 | |
| ChainBackbone=Gemma-3-12B2026.05 | 88.5 | |
| ChainBackbone=GPT-5.4-nano2026.05 | 88.5 | |
| GSIn=12025.06 | 88.3 | |
| Self-Refine + k-way SCBase Model=GPT-3.5-Turbo, k=1202026.03 | 88.1 | |
| S-BoN (s)n=162025.06 | 88.1 | |
| Majority VoteModel=Falcon3-7B2026.03 | 88 | |
| RingBackbone=GPT-5.4-nano2026.05 | 88 | |
| ChainBackbone=Llama-3.1-8B2026.05 | 87.5 | |
| Vote PromptModel=Falcon3-7B2026.03 | 87 | |
| Uncertain PromptModel=Falcon3-7B2026.03 | 86.7 | |
| StarBackbone=Llama-3.1-8B2026.05 | 86.5 | |
| DARModel=Llama3.1-8B2026.03 | 86.3 | |
| RandomBackbone=Llama-3.1-8B2026.05 | 86 | |
| Society Of MindModel=Llama3.1-8B2026.03 | 85.5 |