Tool Use on ToolUse
69.2Task AccuracyGRPO
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| GRPOBackbone=Qwen2.5-7B-Instruct, Training Steps=752026.05 | 69.2 | — | — | — | — | — | — | — | — | |
| AMR-SDBackbone=Qwen3-8B, Training Steps=752026.05 | 67.7 | — | — | — | — | — | — | — | — | |
| GRPOBackbone=Qwen3-8B, Training Steps=752026.05 | 66.9 | — | — | — | — | — | — | — | — | |
| SDPOBackbone=Qwen2.5-7B-Instruct, Training Steps=152026.05 | 66.1 | — | — | — | — | — | — | — | — | |
| AMR-SDBackbone=Qwen2.5-7B-Instruct, Training Steps=752026.05 | 66.1 | — | — | — | — | — | — | — | — | |
| RLSDBackbone=Qwen3-8B, Training Steps=752026.05 | 65.5 | — | — | — | — | — | — | — | — | |
| RLSDBackbone=Qwen3-8B, Training Steps=152026.05 | 64.3 | — | — | — | — | — | — | — | — | |
| SDPOBackbone=Qwen3-8B, Training Steps=152026.05 | 64.2 | — | — | — | — | — | — | — | — | |
| GRPOBackbone=Qwen3-8B, Training Steps=152026.05 | 63.7 | — | — | — | — | — | — | — | — | |
| RLSDBackbone=Qwen2.5-7B-Instruct, Training Steps=752026.05 | 63.6 | — | — | — | — | — | — | — | — | |
| RLSDBackbone=Qwen2.5-7B-Instruct, Training Steps=152026.05 | 63.3 | — | — | — | — | — | — | — | — | |
| AMR-SDBackbone=Qwen3-8B, Training Steps=152026.05 | 63.2 | — | — | — | — | — | — | — | — | |
| GRPOBackbone=Qwen2.5-7B-Instruct, Training Steps=152026.05 | 61.6 | — | — | — | — | — | — | — | — | |
| AMR-SDBackbone=Qwen2.5-7B-Instruct, Training Steps=152026.05 | 61 | — | — | — | — | — | — | — | — | |
| Qwen3-8BBackbone=Qwen3-8B, Training Steps=Untrained2026.05 | 57.3 | — | — | — | — | — | — | — | — | |
| ProFILCondition=ProFIL, Decoding strategy=greedy@1, Theater pooling=chain-level2026.05 | 55.9 | 0.468 | 11 | 14.7 | — | 1,018 | — | — | — | |
| GRPO BaselineCondition=Baseline, Decoding strategy=greedy@1, Theater pooling=chain-level2026.05 | 52.9 | 0.527 | — | 2.9 | 0.976 | 1,248 | — | — | — | |
| Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct, Training Steps=Untrained2026.05 | 52.4 | — | — | — | — | — | — | — | — | |
| BaselineStage=Baseline2026.07 | — | — | — | — | — | — | 48.9 | — | — | |
| GKDDistillation Mode=Teacher-Student2026.05 | — | — | — | — | — | — | 55.91 | 60.05 | — | |
| GRPOStage=Current2026.07 | — | — | — | — | — | — | 58.71 | — | 20 | |
| GRPOStage=Last2026.07 | — | — | — | — | — | — | 57.72 | — | 18 | |
| MOPDDistillation Mode=Teacher-Student2026.05 | — | — | — | — | — | — | 66.44 | 68.61 | — | |
| MOPDDistillation Mode=Self-distillation2026.05 | — | — | — | — | — | — | 64.61 | 66.81 | — | |
| OPDDistillation Mode=Teacher-Student2026.05 | — | — | — | — | — | — | 59.63 | 61.58 | — | |
| Qwen3-14BRole=Teacher2026.05 | — | — | — | — | — | — | 56.02 | 60.89 | — | |
| Qwen3-4BRole=Student2026.05 | — | — | — | — | — | — | 53.27 | 58.58 | — | |
| SDPODistillation Mode=Self-distillation2026.05 | — | — | — | — | — | — | 62.04 | 65.97 | — | |
| SDPO0Stage=Current2026.07 | — | — | — | — | — | — | 37.35 | — | 24 | |
| SDPO0Stage=Last2026.07 | — | — | — | — | — | — | 9.93 | — | 80 | |
| SDPO5%Stage=Current2026.07 | — | — | — | — | — | — | 0 | — | 100 | |
| SDPO5%Stage=Last2026.07 | — | — | — | — | — | — | 0 | — | 100 | |
| SDPOCoT0Stage=Current2026.07 | — | — | — | — | — | — | 38.6 | — | 21 | |
| SDPOCoT0Stage=Last2026.07 | — | — | — | — | — | — | 23.35 | — | 52 |