|
--- |
|
license: llama2 |
|
language: |
|
- en |
|
--- |
|
# DAMA |
|
|
|
<!-- Provide a quick summary of what the model is/does. --> |
|
|
|
## Model |
|
|
|
LLaMA model adapted to mitigate gender bias in text generation. |
|
For adaptation, we used **D**ebiasing **A**lgorithm through **M**odel **A**daptation (DAMA) method described in [Limisiewicz et al., 2024](https://openreview.net/pdf?id=XIZEFyVGC9). |
|
|
|
### Model Description |
|
|
|
<!-- Provide a longer summary of what this model is. --> |
|
|
|
|
|
|
|
- **Developed by:** Tomasz Limisiewicz, David Mareček, Tomáš Musil |
|
- **Funded by:** Grant Agency Czech Republic |
|
- **Language(s) (NLP):** English |
|
- **Adapted from model:** LLaMA |
|
|
|
### Model Sizes |
|
|
|
- **[7B](https://huggingface.co/ufal/DAMA-7B)** |
|
- **[13B](https://huggingface.co/ufal/DAMA-13B)** |
|
- **[33B](https://huggingface.co/ufal/DAMA-33B)** |
|
- **[65B](https://huggingface.co/ufal/DAMA-65B)** |
|
|
|
### Model Sources |
|
|
|
<!-- Provide the basic links for the model. --> |
|
|
|
- **[Repository](github.com/tomlimi/DAMA)** |
|
- **[Paper](openreview.net/pdf?id=XIZEFyVGC9)** |
|
|
|
|
|
|
|
## Bias, Risks, and Limitations |
|
|
|
<!-- This section is meant to convey both technical and sociotechnical limitations. --> |
|
|
|
The model mitigates the gender bias of the original model. |
|
It is better suited for generating and processing texts in sensitive domains. |
|
However, we recommend caution for such use cases because the models retain bias. |
|
|
|
|
|
|
|
## Adaptation |
|
|
|
<!-- Include image. --> |
|
|
|
 |
|
|
|
Schema (b) shows DAMA intervention in a LLaMA layer. |
|
Even though `I - P_c` is depicted as a separate module, in practice, it is multiplied with the output matrix of a feed-forward layer (`W_FF`). |
|
Therefore, DAMA is neutral to the model's parameter count and architecture. |
|
(a) We show the behavior of the model when presented with a stereotypical prompt. |
|
Specifically, (c) shows the projections of the feed-forward latent vector (`u`) onto the output space. |
|
With DAMA (lower arrow), we nullify the gender component of the representation. |
|
It results in balanced probabilities of gendered tokens in the model's output, as shown in (d). |
|
|
|
The method for obtaining `P_c` is based on the Partial Least Square algorithm. |
|
For more details, please refer to the [paper](https://openreview.net/pdf?id=XIZEFyVGC9). |
|
|
|
## Evaluation |
|
|
|
We evaluate the models on multiple benchmarks to assess gender bias and language understanding capabilities. |
|
DAMA models are compared with the original LLaMA models. |
|
|
|
|
|
### Bias Evaluation |
|
|
|
We introduced a metric for evaluating gender bias in text generation. |
|
It measures to which extent the models' output is affected by stereotypical `a_s` and factual `a_f` gender signals. |
|
|
|
Moreover, we provide the scores for two established bias benchmarks: **WinoBias** and **Stereoset**. |
|
|
|
### Results |
|
|
|
|
|
| | Bias | in | LM | | WinoBias | | | StereoSet | | |
|
|--------------------------------------------------------------------|--------|-------|--------|--------|-----------|-----------|------|-----------|------| |
|
| | `a_s` | `a_f` | `b` | Acc | `Delta S` | `Delta G` | lms | ss | ICAT | |
|
| LLaMA 7B | 0.235 | 0.320 | 0.072 | 59.1\% | 40.3\% | 3.0\% | 95.5 | 71.9 | 53.7 | |
|
| DAMA 7B | -0.005 | 0.038 | -0.006 | 57.3\% | 31.5\% | 2.3\% | 95.5 | 69.3 | 58.5 | |
|
| LLaMA 13B | 0.270 | 0.351 | 0.070 | 70.5\% | 35.7\% | -1.5\% | 95.2 | 71.4 | 54.4 | |
|
| DAMA 13B | 0.148 | 0.222 | 0.059 | 66.4\% | 31.1\% | -1.1\% | 94.4 | 68.6 | 59.4 | |
|
| LLaMA 33B | 0.265 | 0.343 | 0.092 | 71.0\% | 36.0\% | -4.0\% | 94.7 | 68.4 | 59.9 | |
|
| DAMA 33B | 0.105 | 0.172 | 0.059 | 63.7\% | 26.7\% | -3.7\% | 94.8 | 65.7 | 65.0 | |
|
| LLaMA 65B | 0.249 | 0.316 | 0.095 | 73.3\% | 35.7\% | 1.4\% | 94.9 | 69.5 | 57.9 | |
|
| DAMA 65B | 0.185 | 0.251 | 0.100 | 71.1\% | 27.2\% | 0.8\% | 92.8 | 67.1 | 61.1 | |
|
| Bias evaluation for the LLaMA models and their debiased instances. | | | | | | | | | | |
|
|
|
|
|
### Performance Evaluation |
|
|
|
To check the effect of debiasing on LM capabilities, we compute perplexity on Wikipedia corpus. |
|
We also test performance on four language understanding end-tasks: **OpenBookQA**, **AI2 Reasoning Challenge** (Easy and Chalange Sets), and **Massive Multitask Language Understanding**. |
|
|
|
|
|
### Results |
|
|
|
| | Perpelexity | ARC-C | ARC-E |OBQA | MMLU | |
|
|-----------|----------------|----------------|-----------|-----------------|-------| |
|
| LLaMA 7B | 26.1 | 42.2 |69.1 | 57.2 | 30.3 | |
|
| DAMA 7B | 28.9 | 41.8 | 68.3 | 56.2 | 30.8 | |
|
| LLaMA 13B | 19.8 | 44.9 | 70.6 | 55.4 | 43.3 | |
|
| DAMA 13B | 21.0 | 44.7 | 70.3 | 56.2 | 43.5 | |
|
| LLaMA 33B | 20.5 | 47.4 | 72.9 | 59.2 | 55.7* | |
|
| DAMA 33B | 19.6 | 45.2 | 71.6 | 58.2 | 56.1* | |
|
| LLaMA 65B | 19.5 | 44.5 | 73.9 | 59.6 | ---* | |
|
| DAMA 65B | 20.1 | 40.5 | 67.7 | 57.2 | --- * | |
|
|
|
Performance evaluation for the \llama{} models and their debiased instances. |
|
Due to hardware limitations, we could not run MMLU inference for 65B models. |
|
In the evaluation of 33B model, we excluded 4\% longest prompts. |
|
|
|
## Citation |
|
|
|
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> |
|
|
|
**BibTeX:** |
|
|
|
``` |
|
@inproceedings{ |
|
limisiewicz2024debiasing, |
|
title={Debiasing Algorithm through Model Adaptation}, |
|
author={Tomasz Limisiewicz and David Mare{\v{c}}ek and Tom{\'a}{\v{s}} Musil}, |
|
booktitle={The Twelfth International Conference on Learning Representations}, |
|
year={2024}, |
|
url={https://openreview.net/forum?id=XIZEFyVGC9} |
|
} |
|
``` |
|
|
|
## Model Card Author |
|
|
|
[Tomasz Limisiewicz](mailto:[email protected]) |
