Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time
Paper
•
2203.05482
•
Published
•
7
merge
This is a merge of pre-trained language models created using mergekit.
Merge Details
Merge Method
This model was merged using the Linear merge method.
Models Merged
The following models were included in the merge:
- Quazim0t0/Phi4.Turn.R1Distill_v1.5.1-Tensors
- bunnycore/Phi-4-ReasoningRP
- bunnycore/Phi-4-Model-Stock-v4
Configuration
The following YAML configuration was used to produce this model:
models:
- model: Quazim0t0/Phi4.Turn.R1Distill_v1.5.1-Tensors
parameters:
weight: 1.0
- model: bunnycore/Phi-4-Model-Stock-v4
parameters:
weight: 1.0
- model: bunnycore/Phi-4-ReasoningRP
parameters:
weight: 1.0
merge_method: linear
normalize: false
int8_mask: true
dtype: bfloat16
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 40.95 |
| IFEval (0-Shot) | 65.87 |
| BBH (3-Shot) | 56.11 |
| MATH Lvl 5 (4-Shot) | 47.96 |
| GPQA (0-shot) | 11.63 |
| MuSR (0-shot) | 14.94 |
| MMLU-PRO (5-shot) | 49.21 |
- Downloads last month
- 11
Model tree for bunnycore/Phi-4-RR-Shoup
Merge model
this model
Paper for bunnycore/Phi-4-RR-Shoup
Evaluation results
- strict accuracy on IFEval (0-Shot)Open LLM Leaderboard65.870
- normalized accuracy on BBH (3-Shot)Open LLM Leaderboard56.110
- exact match on MATH Lvl 5 (4-Shot)Open LLM Leaderboard47.960
- acc_norm on GPQA (0-shot)Open LLM Leaderboard11.630
- acc_norm on MuSR (0-shot)Open LLM Leaderboard14.940
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard49.210