Instruction Following on Arena-Hard Vanilla
57.5Instruction Following ScoreIterDPO (via Rubric-ARROW)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| IterDPO (via Rubric-ARROW)Backbone=Qwen2.5-7B-Instruct2026.05 | 57.5 | 53 | |
| RL (via OnlineRubrics)Backbone=Qwen2.5-7B-Instruct2026.05 | 56.5 | — | |
| IterDPO (via Rubric-RM)Backbone=Qwen2.5-7B-Instruct2026.05 | 56.3 | 51.3 | |
| IterDPO (via Rubric-ARROWw/o RL)Backbone=Qwen2.5-7B-Instruct2026.05 | 55.8 | 49.4 | |
| IterDPO (via RIFL)Backbone=Qwen2.5-7B-Instruct2026.05 | 55.2 | 49.7 | |
| DPO (via Skywork)Backbone=Qwen2.5-7B-Instruct2026.05 | 55.1 | 47.9 | |
| DPO (via RLCF)Backbone=Qwen2.5-7B-Instruct2026.05 | 54.6 | 44.1 | |
| DPO (via Rubric-ARROW)Backbone=Qwen2.5-7B-Instruct2026.05 | 53.5 | 50.7 | |
| DPO (via Rubric-RM)Backbone=Qwen2.5-7B-Instruct2026.05 | 52.9 | 48.6 | |
| DPO (via Ultrafbk.)Backbone=Qwen2.5-7B-Instruct2026.05 | 52.8 | 43.3 | |
| DPO (via Rubric-ARROWw/o RL)Backbone=Qwen2.5-7B-Instruct2026.05 | 52.6 | 47.4 | |
| DPO (via RIFL)Backbone=Qwen2.5-7B-Instruct2026.05 | 52.4 | 47.5 | |
| Qwen2.5-7B-Instruct2026.05 | 51.3 | 41 | |
| IterDPO (via RLCF)Backbone=Qwen2.5-7B-Instruct2026.05 | 51.1 | 46 | |
| DPO (via AI Judge)Backbone=Qwen2.5-7B-Instruct2026.05 | 51 | 39.4 | |
| DPO (via ArmoRM)Backbone=Qwen2.5-7B-Instruct2026.05 | 50.8 | 43.2 | |
| GPT-4 (0314)2026.05 | 50 | 39.4 | |
| SFT (Distilled)Backbone=Qwen2.5-7B-Instruct2026.05 | 32.6 | 32.8 | |
| UltraIF2026.05 | 31.4 | — |