Generative Search on Industry dataset Order (test)
23.14HR@10OneSearch-V2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OneSearch-V22026.03 | 23.14 | 11.51 | |
| OneSearch + TPMAmode=sft model + new task only2026.03 | 22.65 | 11.36 | |
| OneSearch + GRPOmode=sft model + new task only2026.03 | 22.48 | 11.06 | |
| OneSearch + PARSmode=sft model + new task only2026.03 | 22.21 | 10.67 | |
| OneSearch + focal lossmode=incremental (former model + new task)2026.03 | 22.14 | 10.48 | |
| OneSearch + FGMmode=incremental (former model + new task)2026.03 | 21.8 | 10.47 | |
| OneSearch + rdropmode=incremental (former model + new task)2026.03 | 21.68 | 10.45 | |
| OneSearch + self-distillmode=incremental (former model + new task)2026.03 | 21.63 | 10.17 | |
| OneSearch + CoT tasksmode=incremental (former model + new task)2026.03 | 20.94 | 10.08 | |
| OneSearch2026.03 | 20.46 | 9.85 |