Instruction Following Evaluation on IFBench
73.2ScoreRubric-ARROW-voting@5
Evaluation Results
| Method | Links | |
|---|---|---|
| Rubric-ARROW-voting@5Voting Strategy=voting@52026.05 | 73.2 | |
| RIFL-voting@5Voting Strategy=voting@52026.05 | 69.8 | |
| Gemini-2.5-FlashModel Category=Black-box LLMs2026.05 | 69.3 | |
| RIFLModel Category=Rubric-based Methods2026.05 | 69.2 | |
| Rubric-ARROW w/o RL-voting@5Voting Strategy=voting@5, Training Protocol=w/o RL2026.05 | 68.5 | |
| Rubric-ARROWModel Category=Rubric-based Methods2026.05 | 68.2 | |
| Rubric-RM-voting@5Voting Strategy=voting@52026.05 | 67.1 | |
| API Prompting (Rubric+Judge pairwise)Evaluation Protocol=Rubric+Judge pairwise2026.05 | 66.2 | |
| Rubric-ARROW w/o RLTraining Protocol=w/o RL2026.05 | 65.5 | |
| Rubric-RMModel Category=Rubric-based Methods2026.05 | 65.4 | |
| RM-R1-14B (Qwen-2.5-Inst)Backbone=Qwen-2.5-Inst2026.05 | 60.8 | |
| RRM-32BModel Category=Larger White-box LLMs2026.05 | 60.8 | |
| API Prompting (direct Judge)Evaluation Protocol=direct Judge2026.05 | 60.4 | |
| RM-R1-32B (Qwen-2.5-Inst)Backbone=Qwen-2.5-Inst2026.05 | 60.4 | |
| RM-R1-14B (DeepSeek-Dist)Backbone=DeepSeek-Dist2026.05 | 59 | |
| RM-R1-32B (DeepSeek-Dist)Backbone=DeepSeek-Dist2026.05 | 58.6 | |
| RM-R1-7B (DeepSeek-Dist)Backbone=DeepSeek-Dist2026.05 | 56.5 | |
| Qwen-3-8B (Rubric+Judge pairwise)Backbone=Qwen-3-8B, Evaluation Protocol=Rubric+Judge pairwise2026.05 | 55.6 | |
| RM-R1-7B (Qwen-2.5-Inst)Backbone=Qwen-2.5-Inst2026.05 | 55.2 | |
| RRM-7BModel Category=White-box Judge/Reward LLMs2026.05 | 53.2 | |
| Qwen-3-8B (Rubric+Judge pointwise)Backbone=Qwen-3-8B, Evaluation Protocol=Rubric+Judge pointwise2026.05 | 50.2 | |
| JudgeLRM-7BModel Category=White-box Judge/Reward LLMs2026.05 | 47.5 | |
| API Prompting (Rubric+Judge pointwise)Evaluation Protocol=Rubric+Judge pointwise2026.05 | 33.6 |