Multiple-choice Question Answering on HLE filtered subset
19.6NCommGroupDebate
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GroupDebateLLM=DeepSeek-V3.12026.05 | 19.6 | 108.7 | 14.9 | |
| S^2-MADLLM=DeepSeek-V3.12026.05 | 18.7 | 71.7 | 16.7 | |
| GroupDebateLLM=GPT-OSS-120B2026.05 | 18.4 | 111 | 20.2 | |
| SID-ETLLM=DeepSeek-V3.12026.05 | 16.6 | 33.7 | 14.1 | |
| S^2-MADLLM=GPT-OSS-120B2026.05 | 14 | 67.7 | 21 | |
| SID-ETLLM=GPT-OSS-120B2026.05 | 12.8 | 39.6 | 17.2 | |
| SVR-MADLLM=GPT-OSS-120B2026.05 | 7.3 | 39.1 | 21.4 | |
| SVR-MADLLM=DeepSeek-V3.12026.05 | 5.3 | 30.3 | 16.7 | |
| Self ConsistencyLLM=GPT-OSS-120B2026.05 | 0 | 14.1 | 13.9 | |
| Self ConsistencyLLM=DeepSeek-V3.12026.05 | 0 | 9.8 | 14.1 |