Deep search on Average webw., hle, gaia
9.87AccuracyQwen3-8B + TEPOdense
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-8B + TEPOdenseTraining Method=TEPOdense2026.02 | 9.87 | 1.84 | |
| Qwen3-8B + AEPOTraining Method=AEPO2026.02 | 9.69 | 3.21 | |
| Qwen3-8B + TEPOsparseTraining Method=TEPOsparse2026.02 | 9.66 | 1.23 | |
| Qwen3-8B + ARPOTraining Method=ARPO2026.02 | 9.48 | 4.63 | |
| Qwen3-8B + GRPOTraining Method=GRPO2026.02 | 9.43 | 1.77 | |
| Qwen3-8B + SFTTraining Method=SFT2026.02 | 9.06 | 5.65 | |
| Qwen3-8BTraining Method=Base2026.02 | 6.27 | 1.14 |