Share your thoughts, 1 month free Claude Pro on usSee more

Mathematics on GSM8K

94.62GSM8K Score

JT-Safe-V2-35B

Updated 8d ago

Evaluation Results

Method	Links
JT-Safe-V2-35B 2026.05		94.62
SOTA with Equivalent Parameters 2026.05		93.93
Qwen2-72B-Instruct 2024.07		93.2
Llama-3-70B-Instruct 2024.07		93
Mi:dm 2.0 Base-inst 2026.01		91.6
SKillWeave (Ours) 2026.05		91
EMR-Merging 2026.05		90.8
SkillWeave (Delta-Come replacement) 2026.05		90.7
Qwen3-4B 2026.01		90.4
Gemma2-27B-it 2026.05		90.4
Qwen2.5-14B 2026.05		90.2
TALL-Mask 2026.05		90.1
SkillWeave (ASVD replacement) 2026.05		89.7
Twin-Merging r1024 2026.05		89.3
SkillWeave (Self-Rewarding replacement) 2026.05		89.2
Mixtral-8x22B-Instruct 2024.07		89.1
PCB-Merge+DARE 2026.05		88.9
Qwen3-14B 2026.01		88
FuseChat3.0 2026.05		88
Routed LoRA r1024 2026.05		87.9
PCB-Merging 2026.05		87.7
Twin-Merging r512 2026.05		87.6
Jointly MTL 2026.05		87.5
SkillWeave (Self-Specialize replacement) 2026.05		87.3
Ties-Merging 2026.05		87.2
Self-MoE 2026.05		87
SkillWeave (PEFT replacement) 2026.05		86.8
Sequentially MTL 2026.05		86.7
Task Arithmetic 2026.05		86.4
Routed LoRA r512 2026.05		86.4
FuseLLM 2026.05		85.6
Self-Align 2026.05		85.5
Qwen2-57BA14B-it 2026.05		85.3
Qwen1.5-110B-Chat 2024.07		84.5
Llama3.1-8B-Instruct 2026.05		84.5
Self-Rewarding 2026.05		84.3
Mi:dm 2.0 Mini-inst 2026.01		83.1
Full Attention 2026.04		82.75
Qwen1.5-72B-Chat 2024.07		82.7
Qwen1.5-72B-Chat 2026.05		82.7
Exaone-3.5-2.4B-inst 2026.01		82.5
Hybrid-SWA 2026.04		81.92
Llama-3.1-8B-inst 2026.01		81.2
Exaone-3.5-7.8B-inst 2026.01		81.1
KSA 2026.04		81.09
Hybrid-SCA 2026.04		80.1
Hybrid-KSA 2026.04		79.5
Hybrid-Linear 2026.04		72.44
UM-190k 2025.11		62.15
Qwen2-1.5B 2024.07		61.6
UM-187k 2025.11		61.29
UltraFB 2025.11		60.8
UM-170k 2025.11		60.1
TuluDPO 2025.11		59.97
Hybrid-KSA 2026.04		59.14
Qwen2-1.5B 2024.07		58.5
FuseChat3.0 2026.05		57.4
Phi-2 2024.07		57.2
ORPO 2025.11		57.16
SkillWeave 2026.05		56.7
HelpSteer 2025.11		55.27
KSA 2026.04		54.81
CodePref 2025.11		54.28
self-rewarding 2026.05		54.2
Twin-merging 2026.05		53.4
SFT 2025.11		53.07
Hybrid-SCA 2026.04		52.39
Hybrid-GDN 2026.04		50.95
UM-190k 2025.11		50.14
UM-187k 2025.11		49.45
ORPO 2025.11		48.98
TuluDPO 2025.11		48.84
Full 2026.04		48.29
HelpSteer 2025.11		48.07
UltraFB 2025.11		47.99
UM-170k 2025.11		47.96
Hybrid-SWA 2026.04		47.46
SFT 2025.11		46.84
PEFT 2026.05		46.1
Llama3.2-1B-Instruct 2026.05		45.8
CodePref 2025.11		44.98
Qwen2-0.5B 2024.07		40.1
Qwen1.5-1.8B 2024.07		38.4
Qwen2-0.5B 2024.07		36.5
Qwen1.5-1.8B 2024.07		35.3
Gemma-2B 2024.07		17.7
Qwen1.5-0.5B 2024.07		11.3