Math Word Problem solving on SVAMP
94.5Value AccuracyNPG-Muse-8B
Evaluation results
| Method | Links | |
|---|---|---|
| NPG-Muse-8BBackbone=Qwen3-8B-Base2025.08 | 94.5 | |
| Qwen2.5-14B-Instruct-1MSize=14B, Type=Instruct, Context Length=1M2025.08 | 94.3 | |
| NPG-Muse-7BBackbone=Qwen2.5-7B-Instruct-1M2025.08 | 93.6 | |
| Qwen2.5-7B-Instruct-1MSize=7B, Type=Instruct, Context Length=1M2025.08 | 93 | |
| Qwen3-14B-BaseSize=14B, Type=Base2025.08 | 92.3 | |
| RAPSNumber of agents=52026.02 | 92.2 | |
| G-DesignerNumber of agents=52026.02 | 90.7 | |
| AFlowNumber of agents=52026.02 | 90 | |
| LLM-DebateNumber of agents=52026.02 | 89.8 | |
| AgentPruneNumber of agents=52026.02 | 89.8 | |
| PuppeteerNumber of agents=52026.02 | 89.5 | |
| MaASNumber of agents=52026.02 | 89.3 | |
| ComplexCoTNumber of agents=12026.02 | 89.2 | |
| SCNumber of agents=12026.02 | 88.9 | |
| TreeNumber of agents=52026.02 | 88.5 | |
| GPTSwarmNumber of agents=52026.02 | 88.5 | |
| CoTNumber of agents=12026.02 | 88.4 | |
| LLM-BlenderNumber of agents=52026.02 | 88.3 | |
| StarNumber of agents=52026.02 | 88.2 | |
| Vanilla IONumber of agents=1, Backbone=GPT-4o-mini2026.02 | 87.5 | |
| MAS-ZeroNumber of agents=52026.02 | 87.3 | |
| RandomNumber of agents=52026.02 | 87 | |
| AutoAgentsNumber of agents=52026.02 | 86.7 | |
| Qwen3-8B-BaseSize=8B, Type=Base2025.08 | 83.3 | |
| ChainNumber of agents=52026.02 | 82.6 | |
| MI-based distillationBackbone=T5-base2024.03 | 68 | |
| DSSBackbone=T5-base2024.03 | 65.5 | |
| DeBERTa5-fold cross validation=true2023.06 | 63.5 | |
| Single-taskBackbone=T5-base2024.03 | 63 | |
| FinetuningBackbone=T5-base2024.03 | 62.63 | |
| RoBERTa5-fold cross validation=true2023.06 | 30.3 | |
| RNNVAE5-fold cross validation=true2023.06 | 25.9 | |
| RNNEncDec5-fold cross validation=true2023.06 | 25.4 | |
| BERT5-fold cross validation=true2023.06 | 24.8 | |
| DNS5-fold cross validation=true2023.06 | 22.1 | |
| Math-EN5-fold cross validation=true2023.06 | 21.8 | |
| Transformer5-fold cross validation=true2023.06 | 20.7 | |
| GROUP-ATT5-fold cross validation=true2023.06 | 19.2 |