Towards Safety and Helpfulness Balanced Responses via Controllable Large Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914737297555456 |
|---|---|
| author | Tuan, Yi-Lin Chen, Xilun Smith, Eric Michael Martin, Louis Batra, Soumya Celikyilmaz, Asli Wang, William Yang Bikel, Daniel M. |
| author_facet | Tuan, Yi-Lin Chen, Xilun Smith, Eric Michael Martin, Louis Batra, Soumya Celikyilmaz, Asli Wang, William Yang Bikel, Daniel M. |
| contents | As large language models (LLMs) become easily accessible nowadays, the trade-off between safety and helpfulness can significantly impact user experience. A model that prioritizes safety will cause users to feel less engaged and assisted while prioritizing helpfulness will potentially cause harm. Possible harms include teaching people how to build a bomb, exposing youth to inappropriate content, and hurting users' mental health. In this work, we propose to balance safety and helpfulness in diverse use cases by controlling both attributes in LLM. We explore training-free and fine-tuning methods that do not require extra human annotations and analyze the challenges of controlling safety and helpfulness in LLMs. Our experiments demonstrate that our method can rewind a learned model and unlock its controllability. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_01295 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Towards Safety and Helpfulness Balanced Responses via Controllable Large Language Models Tuan, Yi-Lin Chen, Xilun Smith, Eric Michael Martin, Louis Batra, Soumya Celikyilmaz, Asli Wang, William Yang Bikel, Daniel M. Computation and Language Artificial Intelligence As large language models (LLMs) become easily accessible nowadays, the trade-off between safety and helpfulness can significantly impact user experience. A model that prioritizes safety will cause users to feel less engaged and assisted while prioritizing helpfulness will potentially cause harm. Possible harms include teaching people how to build a bomb, exposing youth to inappropriate content, and hurting users' mental health. In this work, we propose to balance safety and helpfulness in diverse use cases by controlling both attributes in LLM. We explore training-free and fine-tuning methods that do not require extra human annotations and analyze the challenges of controlling safety and helpfulness in LLMs. Our experiments demonstrate that our method can rewind a learned model and unlock its controllability. |
| title | Towards Safety and Helpfulness Balanced Responses via Controllable Large Language Models |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2404.01295 |