The Art of Saying No: Contextual Noncompliance in Language Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866915030446899200 |
|---|---|
| author | Brahman, Faeze Kumar, Sachin Balachandran, Vidhisha Dasigi, Pradeep Pyatkin, Valentina Ravichander, Abhilasha Wiegreffe, Sarah Dziri, Nouha Chandu, Khyathi Hessel, Jack Tsvetkov, Yulia Smith, Noah A. Choi, Yejin Hajishirzi, Hannaneh |
| author_facet | Brahman, Faeze Kumar, Sachin Balachandran, Vidhisha Dasigi, Pradeep Pyatkin, Valentina Ravichander, Abhilasha Wiegreffe, Sarah Dziri, Nouha Chandu, Khyathi Hessel, Jack Tsvetkov, Yulia Smith, Noah A. Choi, Yejin Hajishirzi, Hannaneh |
| contents | Chat-based language models are designed to be helpful, yet they should not comply with every user request. While most existing work primarily focuses on refusal of "unsafe" queries, we posit that the scope of noncompliance should be broadened. We introduce a comprehensive taxonomy of contextual noncompliance describing when and how models should not comply with user requests. Our taxonomy spans a wide range of categories including incomplete, unsupported, indeterminate, and humanizing requests (in addition to unsafe requests). To test noncompliance capabilities of language models, we use this taxonomy to develop a new evaluation suite of 1000 noncompliance prompts. We find that most existing models show significantly high compliance rates in certain previously understudied categories with models like GPT-4 incorrectly complying with as many as 30% of requests. To address these gaps, we explore different training strategies using a synthetically-generated training set of requests and expected noncompliant responses. Our experiments demonstrate that while direct finetuning of instruction-tuned models can lead to both over-refusal and a decline in general capabilities, using parameter efficient methods like low rank adapters helps to strike a good balance between appropriate noncompliance and other capabilities. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_12043 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | The Art of Saying No: Contextual Noncompliance in Language Models Brahman, Faeze Kumar, Sachin Balachandran, Vidhisha Dasigi, Pradeep Pyatkin, Valentina Ravichander, Abhilasha Wiegreffe, Sarah Dziri, Nouha Chandu, Khyathi Hessel, Jack Tsvetkov, Yulia Smith, Noah A. Choi, Yejin Hajishirzi, Hannaneh Computation and Language Artificial Intelligence Human-Computer Interaction Chat-based language models are designed to be helpful, yet they should not comply with every user request. While most existing work primarily focuses on refusal of "unsafe" queries, we posit that the scope of noncompliance should be broadened. We introduce a comprehensive taxonomy of contextual noncompliance describing when and how models should not comply with user requests. Our taxonomy spans a wide range of categories including incomplete, unsupported, indeterminate, and humanizing requests (in addition to unsafe requests). To test noncompliance capabilities of language models, we use this taxonomy to develop a new evaluation suite of 1000 noncompliance prompts. We find that most existing models show significantly high compliance rates in certain previously understudied categories with models like GPT-4 incorrectly complying with as many as 30% of requests. To address these gaps, we explore different training strategies using a synthetically-generated training set of requests and expected noncompliant responses. Our experiments demonstrate that while direct finetuning of instruction-tuned models can lead to both over-refusal and a decline in general capabilities, using parameter efficient methods like low rank adapters helps to strike a good balance between appropriate noncompliance and other capabilities. |
| title | The Art of Saying No: Contextual Noncompliance in Language Models |
| topic | Computation and Language Artificial Intelligence Human-Computer Interaction |
| url | https://arxiv.org/abs/2407.12043 |