language:
- en
license: llama3.1
tags:
- medit-mesh
base_model:
- meta-llama/Llama-3.1-8B-Instruct
- arcee-ai/Llama-3.1-SuperNova-Lite
pipeline_tag: text-generation
model-index:
- name: Llama-3.1-MedIT-SUN-8B
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: IFEval (0-Shot)
type: HuggingFaceH4/ifeval
args:
num_few_shot: 0
metrics:
- type: inst_level_strict_acc and prompt_level_strict_acc
value: 78.37
name: strict accuracy
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=meditsolutions/Llama-3.1-MedIT-SUN-8B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: BBH (3-Shot)
type: BBH
args:
num_few_shot: 3
metrics:
- type: acc_norm
value: 32
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=meditsolutions/Llama-3.1-MedIT-SUN-8B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MATH Lvl 5 (4-Shot)
type: hendrycks/competition_math
args:
num_few_shot: 4
metrics:
- type: exact_match
value: 20.02
name: exact match
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=meditsolutions/Llama-3.1-MedIT-SUN-8B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GPQA (0-shot)
type: Idavidrein/gpqa
args:
num_few_shot: 0
metrics:
- type: acc_norm
value: 7.83
name: acc_norm
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=meditsolutions/Llama-3.1-MedIT-SUN-8B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MuSR (0-shot)
type: TAUR-Lab/MuSR
args:
num_few_shot: 0
metrics:
- type: acc_norm
value: 9.64
name: acc_norm
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=meditsolutions/Llama-3.1-MedIT-SUN-8B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU-PRO (5-shot)
type: TIGER-Lab/MMLU-Pro
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 32.4
name: accuracy
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=meditsolutions/Llama-3.1-MedIT-SUN-8B
name: Open LLM Leaderboard
Llama-3.1-MedIT-SUN-8B
Model Description
Llama-3.1-MedIT-SUN-8B is an experimental language model that leverages model merging techniques to combine the capabilities of multiple foundation models. This 8B parameter model is built upon the Llama-3.1-8B-Instruct architecture and represents an exploration in model fusion methodologies.
Key Features
- Base Architecture: Meta's Llama-3.1-8B-Instruct
- Parameter Count: 8 billion
- Development: Created by MedIT Solutions
- Merged Components:
- arcee-ai/Llama-3.1-SuperNova-Lite
- meta-llama/Llama-3.1-8B-Instruct
Technical Details
The model utilizes the proprietary MedIT-mesh technique for model merging, demonstrating an experimental approach to combining language models. This implementation serves as a proof of concept and testing ground for model fusion methodologies.
Purpose
This model was developed primarily for testing and research purposes, exploring the potential of model merging techniques in language model development. It should be considered an experimental release rather than a production-ready model.
Usage Notes
As this is a test model, it is recommended for research and experimental purposes only. Users should be aware of its experimental nature when considering it for any applications.
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 30.04 |
| IFEval (0-Shot) | 78.37 |
| BBH (3-Shot) | 32.00 |
| MATH Lvl 5 (4-Shot) | 20.02 |
| GPQA (0-shot) | 7.83 |
| MuSR (0-shot) | 9.64 |
| MMLU-PRO (5-shot) | 32.40 |