Dialogue on IFEval
80.2IFEvalFuseChat3.0
Evaluation Results
| Method | Links | |
|---|---|---|
| FuseChat3.0#Params=8B2026.05 | 80.2 | |
| SKillWeave (Ours)#Params=10B2026.05 | 79.1 | |
| SkillWeave (Delta-Come replacement)#Params=10B2026.05 | 79 | |
| SkillWeave (Self-Rewarding replacement)#Params=10B2026.05 | 78.8 | |
| Sequentially MTL#Params=8B2026.05 | 78.6 | |
| Twin-Merging r1024#Params=21B2026.05 | 78.4 | |
| SkillWeave (ASVD replacement)#Params=10B2026.05 | 78.3 | |
| Self-Align#Params=8B2026.05 | 78.2 | |
| SkillWeave (PEFT replacement)#Params=10B2026.05 | 78.1 | |
| Qwen2-57BA14B-it#Params=52B2026.05 | 78 | |
| Self-MoE#Params=9B2026.05 | 77.9 | |
| EMR-Merging#Params=16.7B2026.05 | 77.9 | |
| Self-Rewarding#Params=8B2026.05 | 77.9 | |
| TALL-Mask#Params=16.7B2026.05 | 77.8 | |
| Jointly MTL#Params=8B2026.05 | 77.8 | |
| Twin-Merging r512#Params=14.1B2026.05 | 77.7 | |
| Routed LoRA r1024#Params=21B2026.05 | 77.6 | |
| PCB-Merge+DARE#Params=8B2026.05 | 77.2 | |
| Qwen1.5-72B-Chat#Params=72B2026.05 | 77.1 | |
| Gemma2-27B-it#Params=27B2026.05 | 77.1 | |
| PCB-Merging#Params=8B2026.05 | 76.9 | |
| FuseLLM#Params=8B2026.05 | 76.9 | |
| Ties-Merging#Params=8B2026.05 | 76.7 | |
| Routed LoRA r512#Params=14.1B2026.05 | 76.7 | |
| SkillWeave (Self-Specialize replacement)#Params=10B2026.05 | 76.6 | |
| Task Arithmetic#Params=8B2026.05 | 76.2 | |
| Llama3.1-8B-Instruct#Params=8B2026.05 | 75.9 | |
| FuseChat3.0#Params=1.48B, Backbone=Llama-3.2-1B-Instruct2026.05 | 72.5 | |
| SkillWeave#Params=1.42B, Backbone=Llama-3.2-1B-Instruct2026.05 | 64.1 | |
| self-rewarding#Params=1.15B, Backbone=Llama-3.2-1B-Instruct2026.05 | 63.2 | |
| PEFT#Params=1.42B, Backbone=Llama-3.2-1B-Instruct2026.05 | 62.7 | |
| Twin-merging#Params=1.48B, Backbone=Llama-3.2-1B-Instruct2026.05 | 61.9 | |
| Qwen2.5-14B#Params=14B2026.05 | 59.9 | |
| Llama3.2-1B-Instruct#Params=1.15B, Backbone=Llama-3.2-1B-Instruct2026.05 | 58.4 |