Audience Adapted Transforms (LoRA)
Model Description
This repository contains a PEFT LoRA adapter fine-tuned using Adaption AutoScientist on the Audience Adapted Transforms instruction-tuning dataset generated with Adaption's Adaptive Data platform.
The adapter specializes the base Mixtral 8x7B Instruct model to produce explanations tailored for different audiences while preserving factual accuracy, reasoning quality, and communication effectiveness. Training examples span multiple domains including science, medicine, finance, policy, and business.
Model Details
- Developed by: Ekansh116
- Base Model: mistralai/Mixtral-8x7B-Instruct-v0.1
- Training Platform: Adaption AutoScientist
- Library: PEFT (LoRA)
- Model Type: Instruction-tuned LoRA Adapter
- Language: English
Intended Use
This adapter is designed for:
- Audience-aware text generation
- Instruction following
- Educational explanations
- Technical communication
- Cross-domain reasoning
- Research on adaptive language generation
Training Data
The adapter was trained on the Audience Adapted Transforms dataset created using Adaption's Adaptive Data platform.
Dataset characteristics:
- 7,985 instruction-response pairs
- Multi-domain coverage
- Audience-specific explanations
- Science, business, medicine, finance, and policy
- Remastered dataset with A-grade quality
Evaluation
AutoScientist evaluation demonstrated measurable improvements over the base Mixtral 8x7B Instruct model.
Results
| Metric | Base | Adapted |
|---|
| Dataset Win Rate | 36% | 64% |
| General Win Rate | 40% | 60% |
Files
This repository contains a PEFT LoRA adapter and should be loaded together with the base model:
mistralai/Mixtral-8x7B-Instruct-v0.1
Dataset
The training dataset is publicly available.
- Hugging Face Dataset: Ekansh116/adaption-audience-adapted-transforms
- Kaggle Dataset: https://www.kaggle.com/datasets/ekansh0/adaption-audience-adapted-transforms
Acknowledgements
This model was created using Adaption's Adaptive Data and Adaption AutoScientist platforms.
Special thanks to the Adaption team for providing the tools and infrastructure used to generate the training dataset and fine-tune this LoRA adapter.
The base model was developed by Mistral AI.
