Coding on MBPP
94.16ScoreMinistral-3-R
Evaluation Results
| Method | Links | |
|---|---|---|
| Ministral-3-R# Param.=8B2026.02 | 94.16 | |
| Nemotron-Nano-v2# Param.=9B2026.02 | 93.39 | |
| MiniCPM-4.1# Param.=8B2026.02 | 91.05 | |
| Falcon-H1R# Param.=7B2026.02 | 91.05 | |
| MiniCPM-SALA# Param.=9B2026.02 | 89.11 | |
| Qwen3# Param.=8B2026.02 | 81.32 | |
| SkillWeave#Params=1.42B, Backbone=Llama-3.2-1B-Instruct2026.05 | 49.7 | |
| Twin-merging#Params=1.48B, Backbone=Llama-3.2-1B-Instruct2026.05 | 49.4 | |
| Llama3.2-1B-Instruct#Params=1.15B, Backbone=Llama-3.2-1B-Instruct2026.05 | 49 | |
| PEFT#Params=1.42B, Backbone=Llama-3.2-1B-Instruct2026.05 | 48.2 | |
| FuseChat3.0#Params=1.48B, Backbone=Llama-3.2-1B-Instruct2026.05 | 45.5 | |
| self-rewarding#Params=1.15B, Backbone=Llama-3.2-1B-Instruct2026.05 | 44.6 | |
| Hybrid-KSAsetting=Train-from-scratch2026.04 | 36.4 | |
| KSAsetting=Train-from-scratch2026.04 | 35.8 | |
| Hybrid-GDNsetting=Train-from-scratch2026.04 | 34.8 | |
| Hybrid-SCAsetting=Train-from-scratch2026.04 | 31.6 | |
| Fullsetting=Train-from-scratch2026.04 | 30.6 | |
| Hybrid-SWAsetting=Train-from-scratch2026.04 | 30 | |
| Llama-3-70B-InstructType=Instruction-tuned2024.07 | 0.823 | |
| Qwen2-72B-InstructType=Instruction-tuned2024.07 | 0.802 | |
| Qwen1.5-110B-ChatType=Instruction-tuned2024.07 | 0.764 | |
| Mixtral-8x22B-InstructType=Instruction-tuned2024.07 | 0.759 | |
| Qwen1.5-72B-ChatType=Instruction-tuned2024.07 | 0.719 |