Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2411.19038 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915187116736512 |
|---|---|
| author | Ganon, Ben Zolfi, Alon Hofman, Omer Singh, Inderjeet Kojima, Hisashi Elovici, Yuval Shabtai, Asaf |
| author_facet | Ganon, Ben Zolfi, Alon Hofman, Omer Singh, Inderjeet Kojima, Hisashi Elovici, Yuval Shabtai, Asaf |
| contents | In recent years, large language models (LLMs) have had great success in tasks such as casual conversation, contributing to significant advancements in domains like virtual assistance. However, they often generate responses that are not aligned with human values (e.g., ethical standards, safety), leading to potentially unsafe or inappropriate outputs. While several techniques have been proposed to address this problem, they come with a cost, requiring computationally expensive training or dramatically increasing the inference time. In this paper, we present DIESEL, a lightweight inference-guidance technique that can be seamlessly integrated into any autoregressive LLM to semantically filter undesired concepts from the response. DIESEL can function either as a standalone safeguard or as an additional layer of defense, enhancing response safety by reranking the LLM's proposed tokens based on their similarity to predefined negative concepts in the latent space. Our evaluation demonstrates DIESEL's effectiveness on state-of-the-art conversational models, even in adversarial jailbreaking scenarios that challenge response safety. We also highlight DIESEL's generalization capabilities, showing that it can be used in use cases other than safety, providing general-purpose response filtering. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2411_19038 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DIESEL -- Dynamic Inference-Guidance via Evasion of Semantic Embeddings in LLMs Ganon, Ben Zolfi, Alon Hofman, Omer Singh, Inderjeet Kojima, Hisashi Elovici, Yuval Shabtai, Asaf Computation and Language Machine Learning In recent years, large language models (LLMs) have had great success in tasks such as casual conversation, contributing to significant advancements in domains like virtual assistance. However, they often generate responses that are not aligned with human values (e.g., ethical standards, safety), leading to potentially unsafe or inappropriate outputs. While several techniques have been proposed to address this problem, they come with a cost, requiring computationally expensive training or dramatically increasing the inference time. In this paper, we present DIESEL, a lightweight inference-guidance technique that can be seamlessly integrated into any autoregressive LLM to semantically filter undesired concepts from the response. DIESEL can function either as a standalone safeguard or as an additional layer of defense, enhancing response safety by reranking the LLM's proposed tokens based on their similarity to predefined negative concepts in the latent space. Our evaluation demonstrates DIESEL's effectiveness on state-of-the-art conversational models, even in adversarial jailbreaking scenarios that challenge response safety. We also highlight DIESEL's generalization capabilities, showing that it can be used in use cases other than safety, providing general-purpose response filtering. |
| title | DIESEL -- Dynamic Inference-Guidance via Evasion of Semantic Embeddings in LLMs |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2411.19038 |