Multiple-choice Question Answering on IMO-AnswerBench (full)
22.8NCommS^2-MAD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| S^2-MADLLM=GPT-OSS-120B2026.05 | 22.8 | 131.4 | 42.1 | |
| S^2-MADLLM=DeepSeek-V3.12026.05 | 20.9 | 154.7 | 36.8 | |
| SID-ETLLM=DeepSeek-V3.12026.05 | 20 | 103 | 33.8 | |
| GroupDebateLLM=DeepSeek-V3.12026.05 | 19.8 | 234.6 | 40.1 | |
| GroupDebateLLM=GPT-OSS-120B2026.05 | 19.4 | 164.1 | 41.8 | |
| SID-ETLLM=GPT-OSS-120B2026.05 | 17.5 | 85.5 | 38.7 | |
| SVR-MADLLM=DeepSeek-V3.12026.05 | 8.9 | 91.7 | 41.5 | |
| SVR-MADLLM=GPT-OSS-120B2026.05 | 5.6 | 82.1 | 42.8 | |
| Self ConsistencyLLM=GPT-OSS-120B2026.05 | 0 | 45.4 | 36.7 | |
| Self ConsistencyLLM=DeepSeek-V3.12026.05 | 0 | 20.2 | 31.8 |