license: mit
tags:
- ecology
- sustainability
- ecolinguistics
- dpo
datasets:
- neovalle/H4rmony_dpo
model-index:
- name: H4rmoniousAnthea
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 65.87
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=neovalle/H4rmoniousAnthea
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 84.09
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=neovalle/H4rmoniousAnthea
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 63.67
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=neovalle/H4rmoniousAnthea
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 55.08
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=neovalle/H4rmoniousAnthea
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 76.87
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=neovalle/H4rmoniousAnthea
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 12.96
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=neovalle/H4rmoniousAnthea
name: Open LLM Leaderboard
Model Details
Model Description
This model is based on teknium/OpenHermes-2.5-Mistral-7B, DPO fine-tuned with the H4rmony_dpo dataset. Its completions should be more ecologically aware than the base model.
Developed by: Jorge Vallego
Funded by : Neovalle Ltd.
Shared by : airesearch@neovalle.co.uk
Model type: mistral
Language(s) (NLP): Primarily English
License: MIT
Finetuned from model: teknium/OpenHermes-2.5-Mistral-7B
Methodology: DPO
Uses
Intended as PoC to show the effects of H4rmony_dpo dataset with DPO fine-tuning.
Direct Use
For testing purposes to gain insight in order to help with the continous improvement of the H4rmony_dpo dataset.
Downstream Use
Its direct use in applications is not recommended as this model is under testing for a specific task only (Ecological Alignment) Out-of-Scope Use
Not meant to be used other than testing and evaluation of the H4rmony_dpo dataset and ecological alignment. Bias, Risks, and Limitations
This model might produce biased completions already existing in the base model, and others unintentionally introduced during fine-tuning.
How to Get Started with the Model
It can be loaded and run in a Colab instance with High RAM.
Training Details
Trained using DPO
Training Data
H4rmony Dataset - https://huggingface.co/datasets/neovalle/H4rmony_dpo
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 59.76 |
| AI2 Reasoning Challenge (25-Shot) | 65.87 |
| HellaSwag (10-Shot) | 84.09 |
| MMLU (5-Shot) | 63.67 |
| TruthfulQA (0-shot) | 55.08 |
| Winogrande (5-shot) | 76.87 |
| GSM8k (5-shot) | 12.96 |