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 using bunnycore/Qwen2.5-3B-RP-Mix as a base.
Models Merged
The following models were included in the merge:
Configuration
The following YAML configuration was used to produce this model:
merge_method: linear
dtype: bfloat16
normalize: true
base_model: bunnycore/Qwen2.5-3B-RP-Mix
models:
- model: bunnycore/Qwen2.5-3B-RP-Mix
parameters:
weight: 10
density: 1
- model: prithivMLmods/QwQ-LCoT-3B-Instruct
parameters:
weight: 7
density: 0.8
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 22.11 |
| IFEval (0-Shot) | 60.25 |
| BBH (3-Shot) | 28.50 |
| MATH Lvl 5 (4-Shot) | 0.91 |
| GPQA (0-shot) | 2.24 |
| MuSR (0-shot) | 10.76 |
| MMLU-PRO (5-shot) | 29.99 |
- Downloads last month
- 13
Model tree for bunnycore/QwQen-3B-LCoT
Merge model
this model
Space using bunnycore/QwQen-3B-LCoT 1
Paper for bunnycore/QwQen-3B-LCoT
Evaluation results
- strict accuracy on IFEval (0-Shot)Open LLM Leaderboard60.250
- normalized accuracy on BBH (3-Shot)Open LLM Leaderboard28.500
- exact match on MATH Lvl 5 (4-Shot)Open LLM Leaderboard0.910
- acc_norm on GPQA (0-shot)Open LLM Leaderboard2.240
- acc_norm on MuSR (0-shot)Open LLM Leaderboard10.760
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard29.990