SOTA Dialogue on AlpacaEval 2 and PapersWithCode

64.2AlpacaEval2 Score

FuseChat3.0

Updated 12d ago

Evaluation Results

Method	Links
FuseChat3.0 2026.05		64.2
Gemma2-27B-it 2026.05		58.9
SKillWeave (Ours) 2026.05		52.8
SkillWeave (Delta-Come replacement) 2026.05		52.7
SkillWeave (Self-Rewarding replacement) 2026.05		50.8
Twin-Merging r1024 2026.05		49.6
EMR-Merging 2026.05		48.9
TALL-Mask 2026.05		48.6
SkillWeave (PEFT replacement) 2026.05		47.9
Self-Align 2026.05		47.4
Routed LoRA r1024 2026.05		47.1
SkillWeave (ASVD replacement) 2026.05		47
Twin-Merging r512 2026.05		46.6
Qwen2-57BA14B-it 2026.05		46.4
Self-Rewarding 2026.05		46.4
Routed LoRA r512 2026.05		41
Qwen1.5-72B-Chat 2026.05		40.6
Sequentially MTL 2026.05		39.2
Self-MoE 2026.05		38.8
Jointly MTL 2026.05		37.9
Qwen2.5-14B 2026.05		36.4
SkillWeave (Self-Specialize replacement) 2026.05		35.6
PCB-Merge+DARE 2026.05		35
PCB-Merging 2026.05		34.8
Ties-Merging 2026.05		33.9
FuseLLM 2026.05		32.5
Task Arithmetic 2026.05		29.2
Llama3.1-8B-Instruct 2026.05		28.3
FuseChat3.0 2026.05		27.1
SkillWeave 2026.05		12.5
PEFT 2026.05		12.1
self-rewarding 2026.05		11.9
Twin-merging 2026.05		11.5
Llama3.2-1B-Instruct 2026.05		9.7