Instruction Following on IFBench (IFBench score, Training Reward Imp.)
79.79IFBench ScoreSenseNova-U1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SenseNova-U1Parameter Count=30BA3B, Configuration Variant=Think2026.05 | 79.79 | — | |
| Qwen3.5Parameter Count=35BA3B2026.05 | 70.2 | — | |
| SenseNova-U1Parameter Count=8B, Configuration Variant=Think2026.05 | 67.01 | — | |
| Qwen3.5Parameter Count=9B2026.05 | 64.5 | — | |
| GPT-OSS-120BEvaluation Setting=low2026.06 | 58.3 | — | |
| Ling-2.6-flash2026.06 | 57.4 | — | |
| Nemotron-3-Super 120B-A12BEvaluation Setting=non-reasoning2026.06 | 39.87 | — | |
| RLBackbone=Qwen3-4B-Instruct-2507, S=64, M=22026.04 | 39.46 | 15 | |
| GEPABackbone=Qwen3-4B-Instruct-2507, S=32/32, M=/2026.04 | 39.12 | — | |
| GPT-5.4-miniEvaluation Setting=non-reasoning2026.06 | 38.8 | — | |
| p1Backbone=Qwen3-4B-Instruct-2507, S=16, M=82026.04 | 37.41 | 18 | |
| p1Backbone=Qwen3-4B-Instruct-2507, S=4, M=322026.04 | 35.71 | 25 | |
| p1Backbone=Qwen3-4B-Instruct-2507, S=1, M=1282026.04 | 35.37 | 38 | |
| baseBackbone=Qwen3-4B-Instruct-2507, S=/, M=/2026.04 | 35.03 | — | |
| Qwen3VLParameter Count=30BA3B, Configuration Variant=Think2026.05 | 34.69 | — | |
| RLBackbone=Qwen3-1.7B, S=64, M=22026.04 | 31.97 | 5 | |
| GEPABackbone=Qwen3-1.7B, S=32/32, M=/2026.04 | 30.95 | — | |
| Qwen3VLParameter Count=8B, Configuration Variant=Think2026.05 | 29.93 | — | |
| p1Backbone=Qwen3-1.7B, S=16, M=82026.04 | 29.25 | 9 | |
| p1Backbone=Qwen3-1.7B, S=4, M=322026.04 | 26.53 | 13 | |
| Gemma4Parameter Count=26BA4B2026.05 | 25.51 | — | |
| p1Backbone=Qwen3-1.7B, S=1, M=1282026.04 | 24.83 | 27 | |
| baseBackbone=Qwen3-1.7B, S=/, M=/2026.04 | 24.49 | — |