Coding on MBPP (pass@1 and Overall Avg.)

81Overall Average Score

Gemma2-27B-it

Updated 1d ago

Evaluation Results

Method
Gemma2-27B-it 2026.05	81	-
SkillWeave (Delta-Come replacement) 2026.05	78	-
SKillWeave (Ours) 2026.05	77.8	-
SkillWeave (ASVD replacement) 2026.05	76.9	-
Qwen2.5-14B 2026.05	76.7	-
Twin-Merging r1024 2026.05	76.5	-
Routed LoRA r1024 2026.05	76.4	-
SkillWeave (Self-Rewarding replacement) 2026.05	76.3	-
PCB-Merging 2026.05	76.2	-
PCB-Merge+DARE 2026.05	76.2	-
TALL-Mask 2026.05	76.2	-
Ties-Merging 2026.05	76.1	-
Task Arithmetic 2026.05	75.9	-
Routed LoRA r512 2026.05	75.9	-
Jointly MTL 2026.05	75.8	-
EMR-Merging 2026.05	75.7	-
Twin-Merging r512 2026.05	75.5	-
Llama3.1-8B-Instruct 2026.05	75.4	-
Sequentially MTL 2026.05	75.2	-
Self-Rewarding 2026.05	74.9	-
SkillWeave (PEFT replacement) 2026.05	74.2	-
Self-Align 2026.05	73.3	-
SkillWeave (Self-Specialize replacement) 2026.05	72.3	-
Qwen1.5-72B-Chat 2026.05	71.9	-
FuseChat3.0 2026.05	71.8	-
Self-MoE 2026.05	71.2	-
Qwen2-57BA14B-it 2026.05	70.9	-
FuseLLM 2026.05	70.8	-
Anchored Learning 2026.05	52	66.9
Low-SFT 2026.05	45.6	63.8
STM 2026.05	41.8	64.3
Iter-SFT 2026.05	39.6	56.4
Base 2026.05	39.4	62.7
Self-SFT 2026.05	37.8	53.4
DFT 2026.05	24.2	29.1
KL-SFT 2026.05	20.7	24.3
SFT 2026.05	19.1	1.2
GRPO 2025.09	-	71.84
MAPR 2025.09	-	72.97