Editing Models with Task Arithmetic
Paper
•
2212.04089
•
Published
•
7
my-first-blend
This is a merge of pre-trained language models created using mergekit.
Merge Details
Merge Method
This model was merged using the task arithmetic merge method using mistralai/Mistral-7B-Instruct-v0.2 as a base.
Models Merged
The following models were included in the merge:
- SanjiWatsuki/Kunoichi-DPO-v2-7B
- paulml/NeuralOmniWestBeaglake-7B
Configuration
The following YAML configuration was used to produce this model:
models:
- model: SanjiWatsuki/Kunoichi-DPO-v2-7B
parameters:
weight: 0.4
- model: paulml/NeuralOmniWestBeaglake-7B
parameters:
weight: 0.6
base_model: mistralai/Mistral-7B-Instruct-v0.2
merge_method: task_arithmetic
dtype: bfloat16
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 63.66 |
| AI2 Reasoning Challenge (25-Shot) | 69.37 |
| HellaSwag (10-Shot) | 83.03 |
| MMLU (5-Shot) | 53.91 |
| TruthfulQA (0-shot) | 70.70 |
| Winogrande (5-shot) | 79.32 |
| GSM8k (5-shot) | 25.63 |
- Downloads last month
- 46
Model tree for pandego/my-first-blend
Base model
mistralai/Mistral-7B-Instruct-v0.2Paper for pandego/my-first-blend
Evaluation results
- normalized accuracy on AI2 Reasoning Challenge (25-Shot)test set Open LLM Leaderboard69.370
- normalized accuracy on HellaSwag (10-Shot)validation set Open LLM Leaderboard83.030
- accuracy on MMLU (5-Shot)test set Open LLM Leaderboard53.910
- mc2 on TruthfulQA (0-shot)validation set Open LLM Leaderboard70.700
- accuracy on Winogrande (5-shot)validation set Open LLM Leaderboard79.320
- accuracy on GSM8k (5-shot)test set Open LLM Leaderboard25.630