Consistency Evaluation on DecisionBench
0.2902CR MeanDebate-AHP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Debate-AHPLLM generator=Llama3.1-8b2026.01 | 0.2902 | 0.4309 | 0 | |
| Standard-AHPLLM generator=Llama3.1-8b2026.01 | 0.1693 | 0.7514 | 71.42 | |
| Standard-AHPLLM generator=Llama3.1-70b2026.01 | 0.1414 | 1.3921 | 86.95 | |
| Doc2AHPLLM generator=Llama3.1-70b2026.01 | 0.053 | 0.0957 | 100 | |
| Standard-AHPLLM generator=GPT-5.22026.01 | 0.0434 | 0.3167 | 88.57 | |
| Doc2AHPLLM generator=GPT-5.22026.01 | 0.0326 | 0.0882 | 100 | |
| Doc2AHPLLM generator=Llama3.1-8b2026.01 | 0.0139 | 0.0807 | 100 | |
| Debate-AHPLLM generator=GPT-5.22026.01 | 0.0114 | 0.0335 | 100 | |
| Debate-AHPLLM generator=Llama3.1-70b2026.01 | 0.0065 | 0.0132 | 100 |