Instruction Following on AdvancedIF
71AccuracyBRAID
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5.1 medium2025.12 | 71 | 1.44 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5.1 medium2025.12 | 70 | 1.75 | — | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5 medium2025.12 | 69 | 1.05 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5 medium2025.12 | 66 | 0.9 | — | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5 medium2025.12 | 66 | 1.04 | — | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5.1 medium2025.12 | 66 | 1.63 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5.1 medium2025.12 | 66 | 1.67 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-5.1 medium2025.12 | 65 | 1.44 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5 medium2025.12 | 65 | 1.11 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5.1 medium2025.12 | 65 | 1.55 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5.1 medium2025.12 | 64 | 1.26 | — | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5 mini medium2025.12 | 64 | 7.07 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5 medium2025.12 | 63 | 1.01 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5 medium2025.12 | 62 | 1.04 | — | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5 medium2025.12 | 62 | 1.01 | — | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5.1 medium2025.12 | 61 | 1.4 | — | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5 mini medium2025.12 | 60 | 5.94 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5 mini medium2025.12 | 60 | 6.51 | — | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-5 medium2025.12 | 60 | 1.11 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5 mini medium2025.12 | 60 | 6.6 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5 minimal2025.12 | 58 | 3.53 | — | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-5.1 medium2025.12 | 58 | 1.31 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5 nano medium2025.12 | 57 | 16.23 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5.1 none2025.12 | 57 | 2.53 | — | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5 mini medium2025.12 | 57 | 5.21 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-5 medium2025.12 | 57 | 1.31 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5 minimal2025.12 | 56 | 2.97 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-5.1 none2025.12 | 56 | 2.81 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5 mini minimal2025.12 | 55 | 15.34 | — | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5 mini minimal2025.12 | 55 | 14.7 | — | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5.1 none2025.12 | 55 | 2.52 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-5 mini medium2025.12 | 55 | 5.8 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5 mini medium2025.12 | 55 | 6.38 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5 mini medium2025.12 | 54 | 5.14 | — | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5 nano medium2025.12 | 54 | 14.97 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5 minimal2025.12 | 53 | 3.18 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5 nano medium2025.12 | 53 | 14.23 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5 mini minimal2025.12 | 52 | 14.47 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5 nano medium2025.12 | 52 | 14.57 | — | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5.1 none2025.12 | 51 | 2.62 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-4o2025.12 | 51 | 2.73 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5.1 none2025.12 | 51 | 2.36 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5.1 none2025.12 | 50 | 2.08 | — | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5 nano medium2025.12 | 50 | 13.71 | — | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5.1 none2025.12 | 50 | 1.99 | — | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-5 mini medium2025.12 | 49 | 5.07 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5 minimal2025.12 | 49 | 2.93 | — | — | — | |
| RTT-CSR (Ours)Backbone=Qwen3-4B-Instruct2026.04 | 48.39 | — | 80.77 | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5 mini minimal2025.12 | 48 | 15.54 | — | — | — | |
| RTT-AON (Ours)Backbone=Qwen3-4B-Instruct2026.04 | 47.48 | — | 80.79 | — | — | |
| RL-AONBackbone=Qwen3-4B-Instruct2026.04 | 47.29 | — | 79.69 | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5 nano medium2025.12 | 47 | 11.45 | — | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-5 nano medium2025.12 | 47 | 13.19 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-5 minimal2025.12 | 47 | 3.8 | — | — | — | |
| RL-CSRBackbone=Qwen3-4B-Instruct2026.04 | 46.02 | — | 79.81 | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-4o2025.12 | 46 | 12.76 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-4o2025.12 | 46 | 2.36 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-4o2025.12 | 46 | 2.34 | — | — | — | |
| DPOBackbone=Qwen3-4B-Instruct2026.04 | 45.71 | — | 79.33 | — | — | |
| SFTBackbone=Qwen3-4B-Instruct2026.04 | 45.07 | — | 79.23 | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5 minimal2025.12 | 45 | 3.02 | — | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5 minimal2025.12 | 45 | 2.33 | — | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-5 mini minimal2025.12 | 45 | 13.39 | — | — | — | |
| Qwen3-4B-Instruct†Backbone=Qwen3-4B-Instruct2026.04 | 44.31 | — | 79.17 | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5 nano medium2025.12 | 44 | 11.74 | — | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5 minimal2025.12 | 44 | 2.81 | — | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-5 minimal2025.12 | 44 | 3.29 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5 mini minimal2025.12 | 44 | 11.89 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5.1 none2025.12 | 44 | 2.03 | — | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-5.1 none2025.12 | 43 | 1.89 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-5 mini minimal2025.12 | 42 | 12.29 | — | — | — | |
| BRAIDGenerator=gpt-5 medium, Solver=gpt-5 nano minimal2025.12 | 40 | 61.69 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5 mini minimal2025.12 | 39 | 9.88 | — | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5 mini minimal2025.12 | 39 | 9.87 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-4o2025.12 | 38 | 6.22 | — | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-4o2025.12 | 38 | 3.05 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-5 nano medium2025.12 | 38 | 10.36 | — | — | — | |
| RTT-AON (Ours)Backbone=Llama3.2-3B-Instruct2026.04 | 37.99 | — | 74.07 | — | — | |
| RTT-CSR (Ours)Backbone=Llama3.2-3B-Instruct2026.04 | 37.57 | — | 74.37 | — | — | |
| BRAIDGenerator=gpt-5 mini minimal, Solver=gpt-5 nano minimal2025.12 | 37 | 53.51 | — | — | — | |
| BRAIDGenerator=gpt-5.1 none, Solver=gpt-5 nano minimal2025.12 | 37 | 53.78 | — | — | — | |
| RL-CSRBackbone=Llama3.2-3B-Instruct2026.04 | 36.29 | — | 73.47 | — | — | |
| RL-AONBackbone=Llama3.2-3B-Instruct2026.04 | 35.32 | — | 73.53 | — | — | |
| BRAIDGenerator=gpt-5 minimal, Solver=gpt-5 nano minimal2025.12 | 35 | 53.25 | — | — | — | |
| BRAIDGenerator=gpt-5.1 medium, Solver=gpt-5 nano minimal2025.12 | 35 | 53.27 | — | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-4o2025.12 | 34 | 2.39 | — | — | — | |
| RTT-CSR (Ours)Backbone=Qwen2.5-7B-Instruct2026.04 | 33.74 | — | 73.72 | — | — | |
| Llama3.2-3B-Instruct⋆Backbone=Llama3.2-3B-Instruct2026.04 | 33.37 | — | 72.01 | — | — | |
| SFTBackbone=Llama3.2-3B-Instruct2026.04 | 32.89 | — | 72.18 | — | — | |
| RTT-AON (Ours)Backbone=Qwen2.5-7B-Instruct2026.04 | 32.51 | — | 72.7 | — | — | |
| DPOBackbone=Llama3.2-3B-Instruct2026.04 | 32.28 | — | 71.14 | — | — | |
| RL-AONBackbone=Qwen2.5-7B-Instruct2026.04 | 32.1 | — | 72.33 | — | — | |
| BRAIDGenerator=gpt-5 nano medium, Solver=gpt-4o2025.12 | 32 | 2.17 | — | — | — | |
| BRAIDGenerator=gpt-5 nano minimal, Solver=gpt-4o2025.12 | 32 | 2.17 | — | — | — | |
| BRAIDGenerator=gpt-5 mini medium, Solver=gpt-5 nano minimal2025.12 | 31 | 45.39 | — | — | — | |
| RL-CSRBackbone=Qwen2.5-7B-Instruct2026.04 | 30.91 | — | 72 | — | — | |
| DPOBackbone=Qwen2.5-7B-Instruct2026.04 | 30.82 | — | 70.9 | — | — | |
| SFTBackbone=Qwen2.5-7B-Instruct2026.04 | 29.71 | — | 69.61 | — | — | |
| Qwen2.5-7B-Instruct†Backbone=Qwen2.5-7B-Instruct2026.04 | 29.48 | — | 71.31 | — | — | |
| BRAIDGenerator=gpt-4o, Solver=gpt-5 nano minimal2025.12 | 27 | 51.47 | — | — | — |