Grade School Math Reasoning on GSM8K (Accuracy, Latency)
94.53AccuracyAgentDropout
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AgentDropoutBackbone=Qwen2.5-14B-Instruct, Topology=Star, Training-Free (TF)=false2026.05 | 94.53 | 88.62 | |
| LLM-DebateBackbone=Qwen2.5-14B-Instruct, Topology=Debate, Training-Free (TF)=true2026.05 | 94.24 | 112.34 | |
| AgentDropoutBackbone=Qwen2.5-14B-Instruct, Topology=Layered, Training-Free (TF)=false2026.05 | 94.24 | 95.85 | |
| Vanilla MASBackbone=Qwen2.5-14B-Instruct, Topology=Layered, Training-Free (TF)=true2026.05 | 94.14 | 119.64 | |
| AgentDropoutBackbone=Qwen2.5-14B-Instruct, Topology=Random, Training-Free (TF)=false2026.05 | 94.14 | 85.58 | |
| CONCATBackbone=Qwen2.5-14B-Instruct, Topology=Hybrid, Training-Free (TF)=true2026.05 | 94.1 | 51.19 | |
| Vanilla MASBackbone=Qwen2.5-14B-Instruct, Topology=Star, Training-Free (TF)=true2026.05 | 94.06 | 131.6 | |
| SC-CoTBackbone=Qwen2.5-72B-Instruct, Topology=-, Training-Free (TF)=true2026.05 | 94.06 | 131.48 | |
| AgentDropoutBackbone=Qwen2.5-14B-Instruct, Topology=Chain, Training-Free (TF)=false2026.05 | 94.01 | 82.01 | |
| Vanilla MASBackbone=Qwen2.5-14B-Instruct, Topology=Random, Training-Free (TF)=true2026.05 | 93.98 | 117.23 | |
| Vanilla MASBackbone=Qwen2.5-14B-Instruct, Topology=FullConnected, Training-Free (TF)=true2026.05 | 93.98 | 132.1 | |
| AgentDropoutBackbone=Qwen2.5-14B-Instruct, Topology=FullConnected, Training-Free (TF)=false2026.05 | 93.98 | 93.51 | |
| LLM-DebateBackbone=Qwen2.5-72B-Instruct, Topology=Debate, Training-Free (TF)=true2026.05 | 93.93 | 303.27 | |
| CoTBackbone=Qwen2.5-72B-Instruct, Topology=-, Training-Free (TF)=true2026.05 | 93.85 | 37.26 | |
| AgentDropoutBackbone=Qwen2.5-72B-Instruct, Topology=FullConnected, Training-Free (TF)=false2026.05 | 93.85 | 244.47 | |
| Vanilla MASBackbone=Qwen2.5-72B-Instruct, Topology=Layered, Training-Free (TF)=true2026.05 | 93.78 | 312.79 | |
| AgentDropoutBackbone=Qwen2.5-72B-Instruct, Topology=Star, Training-Free (TF)=false2026.05 | 93.72 | 232.63 | |
| Vanilla MASBackbone=Qwen2.5-14B-Instruct, Topology=Chain, Training-Free (TF)=true2026.05 | 93.57 | 110.68 | |
| Vanilla MASBackbone=Qwen2.5-72B-Instruct, Topology=Star, Training-Free (TF)=true2026.05 | 93.49 | 341.8 | |
| CONCATBackbone=Qwen2.5-72B-Instruct, Topology=Hybrid, Training-Free (TF)=true2026.05 | 93.49 | 141.72 | |
| Vanilla MASBackbone=Qwen2.5-72B-Instruct, Topology=Random, Training-Free (TF)=true2026.05 | 93.44 | 308.52 | |
| AgentDropoutBackbone=Qwen2.5-72B-Instruct, Topology=Chain, Training-Free (TF)=false2026.05 | 93.36 | 228.89 | |
| Vanilla MASBackbone=Qwen2.5-72B-Instruct, Topology=FullConnected, Training-Free (TF)=true2026.05 | 93.28 | 334.57 | |
| AgentDropoutBackbone=Qwen2.5-72B-Instruct, Topology=Layered, Training-Free (TF)=false2026.05 | 93.26 | 243.1 | |
| SC-CoTBackbone=Qwen2.5-14B-Instruct, Topology=-, Training-Free (TF)=true2026.05 | 93.05 | 54 | |
| Vanilla MASBackbone=Qwen2.5-72B-Instruct, Topology=Chain, Training-Free (TF)=true2026.05 | 92.86 | 299.55 | |
| AgentDropoutBackbone=Qwen2.5-72B-Instruct, Topology=Random, Training-Free (TF)=false2026.05 | 92.76 | 231.72 | |
| LLM-DebateBackbone=Llama3-8B-Instruct, Topology=Debate, Training-Free (TF)=true2026.05 | 78.8 | 133.92 | |
| AgentDropoutBackbone=Llama3-8B-Instruct, Topology=Chain, Training-Free (TF)=false2026.05 | 78.57 | 85.37 | |
| CONCATBackbone=Llama3-8B-Instruct, Topology=Hybrid, Training-Free (TF)=true2026.05 | 77.08 | 72.02 | |
| Vanilla MASBackbone=Llama3-8B-Instruct, Topology=Layered, Training-Free (TF)=true2026.05 | 76.35 | 158.44 | |
| AgentDropoutBackbone=Llama3-8B-Instruct, Topology=Random, Training-Free (TF)=false2026.05 | 76.32 | 98.48 | |
| AgentDropoutBackbone=Llama3-8B-Instruct, Topology=Star, Training-Free (TF)=false2026.05 | 76.28 | 107.69 | |
| AgentDropoutBackbone=Llama3-8B-Instruct, Topology=Layered, Training-Free (TF)=false2026.05 | 76.25 | 100.57 | |
| Vanilla MASBackbone=Llama3-8B-Instruct, Topology=FullConnected, Training-Free (TF)=true2026.05 | 75.6 | 149.22 | |
| Vanilla MASBackbone=Llama3-8B-Instruct, Topology=Random, Training-Free (TF)=true2026.05 | 75.57 | 135.14 | |
| Vanilla MASBackbone=Llama3-8B-Instruct, Topology=Star, Training-Free (TF)=true2026.05 | 75.49 | 147.11 | |
| AgentDropoutBackbone=Llama3-8B-Instruct, Topology=FullConnected, Training-Free (TF)=false2026.05 | 75.09 | 118.11 | |
| SC-CoTBackbone=Llama3-8B-Instruct, Topology=-, Training-Free (TF)=true2026.05 | 74.19 | 40.77 | |
| Vanilla MASBackbone=Llama3-8B-Instruct, Topology=Chain, Training-Free (TF)=true2026.05 | 73.7 | 122.11 | |
| CoTBackbone=Qwen2.5-14B-Instruct, Topology=-, Training-Free (TF)=true2026.05 | 67.97 | 13.78 | |
| CoTBackbone=Llama3-8B-Instruct, Topology=-, Training-Free (TF)=true2026.05 | 64.64 | 9.2 |