README.md

---
license: apache-2.0
datasets:
- alexl83/AlpacaDataCleaned
- sahil2801/CodeAlpaca-20k
language:
- en
library_name: transformers
pipeline_tag: text-generation
tags:
- rlhf
- alignment
- simulation
- computational social science
---


# Model Card for So(cially)-Good LM

![model image](https://agwarbliu.s3.amazonaws.com/logo.png)

![model image](https://agwarbliu.s3.amazonaws.com/model_select_base.png)


**Fast, Effective, and Stable alternative of RLHF!**

**Instead of training an additional reward model that is likely to be gamed, we directly train the model on the social games!** 🕹️ 🎲 🎮

Full details on simulation and training can be found [here](https://github.com/agi-templar/Stable-Alignment).

# Training Procedure

This is the very beginning of Stable Alignment project, which is an enhanced instruction tuning model based on LLaMA.

We improve:

- Instruction tuning data quality, by using [AlpacaDataCleaned](https://github.com/gururise/AlpacaDataCleaned), which fixes many errors in original Alpaca dataset.

- Code pretraining with [codealpaca](https://github.com/sahil280114/codealpaca).

We use the [Alpaca fine-tuning script](https://github.com/tatsu-lab/stanford_alpaca) to train this model.


# Bias, Risks, and Limitations

Although this project aims to better align current LMs with social norms, inappropriate content and inherent biases in the training data will still impair the alignment of the model.

The model should not be used directly in any application, without a prior assessment of safety and fairness concerns specific to the application.