Graduate-Level Question Answering on GPQA (Acceptance length, Delta (%))
Evaluation results
| Method | Links | | |
|---|
| TTSModel=Llama3.1-8B2026.05 | | 2 | 60.4 |
|---|
| TTSModel=Qwen/Qwen3-8B2026.05 | | 1.9 | 24.5 |
|---|
| TTSModel=Qwen/Qwen3-32B2026.05 | | 1.8 | 21.5 |
|---|
| EAGLE-3Model=Qwen/Qwen3-8B2026.05 | | 1.5 | — |
|---|
| EAGLE-3Model=Qwen/Qwen3-32B2026.05 | | 1.4 | — |
|---|
| EAGLE-3Model=Llama3.1-8B2026.05 | | 1.2 | — |
|---|