Instruction Following on FollowBench (test)
85.1CSRCRGC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CRGCBase Model=Claude-3.7-Sonnet, Model Type=Closed-source2026.06 | 85.1 | 88.6 | 86.3 | |
| CRGCBase Model=GPT-4o, Model Type=Closed-source2026.06 | 83.8 | 87.4 | 84.9 | |
| CRGCBase Model=Gemini-2.5-Pro, Model Type=Closed-source2026.06 | 83 | 86.7 | 84.2 | |
| CRGCBase Model=Llama-3.1-72B, Model Type=Open-source2026.06 | 81.2 | 84.9 | 82.3 | |
| CRGCBase Model=Qwen2.5-72B, Model Type=Open-source2026.06 | 80.4 | 84.2 | 81.5 | |
| Self-Reflection (SR)Base Model=Claude-3.7-Sonnet, Model Type=Closed-source2026.06 | 80.2 | 84 | 81.4 | |
| Self-Reflection (SR)Base Model=GPT-4o, Model Type=Closed-source2026.06 | 78.9 | 82.7 | 80.1 | |
| Chain-of-Thoughts (CoT)Base Model=Claude-3.7-Sonnet, Model Type=Closed-source2026.06 | 78.4 | 82.1 | 79.5 | |
| CRGCBase Model=Mixtral-8x22B, Model Type=Open-source2026.06 | 78.3 | 82 | 79.4 | |
| Self-Reflection (SR)Base Model=Gemini-2.5-Pro, Model Type=Closed-source2026.06 | 78.2 | 82 | 79.5 | |
| Chain-of-Thoughts (CoT)Base Model=GPT-4o, Model Type=Closed-source2026.06 | 77.1 | 80.9 | 78.3 | |
| Chain-of-Thoughts (CoT)Base Model=Gemini-2.5-Pro, Model Type=Closed-source2026.06 | 76.5 | 80.3 | 77.8 | |
| Self-Reflection (SR)Base Model=Llama-3.1-72B, Model Type=Open-source2026.06 | 75.6 | 79.6 | 76.8 | |
| Self-Reflection (SR)Base Model=Qwen2.5-72B, Model Type=Open-source2026.06 | 74.8 | 78.9 | 76 | |
| Standard Prompting (SP)Base Model=Claude-3.7-Sonnet, Model Type=Closed-source2026.06 | 74.3 | 78.2 | 75.7 | |
| Chain-of-Thoughts (CoT)Base Model=Llama-3.1-72B, Model Type=Open-source2026.06 | 73.8 | 77.9 | 75 | |
| Standard Prompting (SP)Base Model=GPT-4o, Model Type=Closed-source2026.06 | 73.2 | 77.3 | 74.5 | |
| Chain-of-Thoughts (CoT)Base Model=Qwen2.5-72B, Model Type=Open-source2026.06 | 72.9 | 77.1 | 74.2 | |
| Standard Prompting (SP)Base Model=Gemini-2.5-Pro, Model Type=Closed-source2026.06 | 72.7 | 76.8 | 74 | |
| Self-Reflection (SR)Base Model=Mixtral-8x22B, Model Type=Open-source2026.06 | 72.5 | 76.6 | 73.8 | |
| Chain-of-Thoughts (CoT)Base Model=Mixtral-8x22B, Model Type=Open-source2026.06 | 70.8 | 74.9 | 72.1 | |
| Standard Prompting (SP)Base Model=Llama-3.1-72B, Model Type=Open-source2026.06 | 70 | 74.2 | 71.4 | |
| Standard Prompting (SP)Base Model=Qwen2.5-72B, Model Type=Open-source2026.06 | 69.2 | 73.5 | 70.6 | |
| Standard Prompting (SP)Base Model=Mixtral-8x22B, Model Type=Open-source2026.06 | 67.2 | 71.4 | 68.5 |