Reasoning on GPQA (80% test)
54.22AccuracySelf-Refine
Evaluation Results
| Method | Links | |
|---|---|---|
| Self-RefineLLM Backbone=Llama3.3-70B2025.05 | 54.22 | |
| ADASLLM Backbone=Llama3.3-70B2025.05 | 53.6 | |
| MADLLM Backbone=Llama3.3-70B2025.05 | 52.61 | |
| MAS-ZEROLLM Backbone=Llama3.3-70B2025.05 | 52.41 | |
| MADLLM Backbone=GPT-4o2025.05 | 52.01 | |
| MAS-ZEROLLM Backbone=Qwen2.5-32B2025.05 | 51.81 | |
| CoT-SCLLM Backbone=Llama3.3-70B2025.05 | 51.2 | |
| MAS-ZEROLLM Backbone=GPT-4o2025.05 | 50.6 | |
| CoTLLM Backbone=Llama3.3-70B2025.05 | 50.6 | |
| DebateLLM Backbone=Llama3.3-70B2025.05 | 50.6 | |
| Self-RefineLLM Backbone=Qwen2.5-32B2025.05 | 50.6 | |
| CoTLLM Backbone=Qwen2.5-32B2025.05 | 50 | |
| SPPLLM Backbone=GPT-4o2025.05 | 49.8 | |
| CoT-SCLLM Backbone=Qwen2.5-32B2025.05 | 49.4 | |
| DebateLLM Backbone=Qwen2.5-32B2025.05 | 49.4 | |
| ReConcileLLM Backbone=GPT-4o2025.05 | 48.43 | |
| AFlowLLM Backbone=Llama3.3-70B2025.05 | 47.59 | |
| ReConcileLLM Backbone=Llama3.3-70B2025.05 | 47.17 | |
| ReConcileLLM Backbone=Qwen2.5-32B2025.05 | 47.17 | |
| ADASLLM Backbone=Qwen2.5-32B2025.05 | 47 | |
| DebateLLM Backbone=GPT-4o2025.05 | 46.99 | |
| DyLANLLM Backbone=GPT-4o2025.05 | 46.99 | |
| AFlowLLM Backbone=GPT-4o2025.05 | 46.99 | |
| MaASLLM Backbone=Qwen2.5-32B2025.05 | 46.99 | |
| Self-RefineLLM Backbone=GPT-4o2025.05 | 46.39 | |
| AFlowLLM Backbone=Qwen2.5-32B2025.05 | 46.39 | |
| SPPLLM Backbone=Llama3.3-70B2025.05 | 46.18 | |
| CoTLLM Backbone=GPT-4o2025.05 | 45.78 | |
| ADASLLM Backbone=GPT-4o2025.05 | 45.2 | |
| MAS-GPTLLM Backbone=GPT-4o2025.05 | 43.98 | |
| MaASLLM Backbone=Llama3.3-70B2025.05 | 43.98 | |
| MADLLM Backbone=Qwen2.5-32B2025.05 | 43.57 | |
| CoT-SCLLM Backbone=GPT-4o2025.05 | 43.37 | |
| MaASLLM Backbone=GPT-4o2025.05 | 43.37 | |
| SPPLLM Backbone=Qwen2.5-32B2025.05 | 42.77 | |
| DyLANLLM Backbone=Qwen2.5-32B2025.05 | 42.57 | |
| DyLANLLM Backbone=Llama3.3-70B2025.05 | 41.57 | |
| MAS-GPTLLM Backbone=Llama3.3-70B2025.05 | 40 | |
| MAS-GPTLLM Backbone=Qwen2.5-32B2025.05 | 37.35 |