SOTA Coding on HumanEval (HumanEval metric) and PapersWithCode

79.9HumanEval

Qwen2-57BA14B-it

Updated 12d ago

Evaluation Results

Method	Links
Qwen2-57BA14B-it 2026.05		79.9
Gemma2-27B-it 2026.05		78.7
SKillWeave (Ours) 2026.05		75
SkillWeave (Delta-Come replacement) 2026.05		74.9
Twin-Merging r1024 2026.05		73.8
SkillWeave (ASVD replacement) 2026.05		73.7
EMR-Merging 2026.05		73.5
SkillWeave (PEFT replacement) 2026.05		73.5
Routed LoRA r1024 2026.05		73.2
Twin-Merging r512 2026.05		73.2
FuseLLM 2026.05		73.2
TALL-Mask 2026.05		72.8
Routed LoRA r512 2026.05		72.5
PCB-Merge+DARE 2026.05		71.5
Qwen1.5-72B-Chat 2026.05		71.3
PCB-Merging 2026.05		71.3
FuseChat3.0 2026.05		71.3
Ties-Merging 2026.05		71.1
Jointly MTL 2026.05		70.7
SkillWeave (Self-Specialize replacement) 2026.05		70.7
Task Arithmetic 2026.05		70.6
Self-MoE 2026.05		70.6
SkillWeave (Self-Rewarding replacement) 2026.05		70.3
Llama3.1-8B-Instruct 2026.05		69.5
Sequentially MTL 2026.05		69.4
Self-Align 2026.05		69.2
Self-Rewarding 2026.05		68.8
Qwen2.5-14B 2026.05		56.7