Adversarial Contrastive Decoding: Boosting Safety Alignment of Large Language Models via Opposite Prompt Optimization

• URL: http://arxiv.org/abs/2406.16743v1
• Date: Mon, 24 Jun 2024 15:51:30 GMT
• Title: Adversarial Contrastive Decoding: Boosting Safety Alignment of Large Language Models via Opposite Prompt Optimization
• Authors: Zhengyue Zhao, Xiaoyun Zhang, Kaidi Xu, Xing Hu, Rui Zhang, Zidong Du, Qi Guo, Yunji Chen,
• Abstract summary: Adversarial Contrastive Decoding (ACD) is an optimization-based framework to generate two opposite system prompts for prompt-based contrastive decoding. ACD achieves much better safety performance than previous model training-free decoding methods without sacrificing original generation ability.
• Score: 34.29833630422768
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: With the widespread application of Large Language Models (LLMs), it has become a significant concern to ensure their safety and prevent harmful responses. While current safe-alignment methods based on instruction fine-tuning and Reinforcement Learning from Human Feedback (RLHF) can effectively reduce harmful responses from LLMs, they often require high-quality datasets and heavy computational overhead during model training. Another way to align language models is to modify the logit of tokens in model outputs without heavy training. Recent studies have shown that contrastive decoding can enhance the performance of language models by reducing the likelihood of confused tokens. However, these methods require the manual selection of contrastive models or instruction templates. To this end, we propose Adversarial Contrastive Decoding (ACD), an optimization-based framework to generate two opposite system prompts for prompt-based contrastive decoding. ACD only needs to apply a lightweight prompt tuning on a rather small anchor dataset (< 3 min for each model) without training the target model. Experiments conducted on extensive models and benchmarks demonstrate that the proposed method achieves much better safety performance than previous model training-free decoding methods without sacrificing its original generation ability.

Related papers

• DSTC: Direct Preference Learning with Only Self-Generated Tests and Code to Improve Code LMs [56.24431208419858]
  We introduce underlinetextbfDirect Preference Learning with Only underlinetextbfSelf-Generated underlinetextbfTests and underlinetextbfCode (DSTC) DSTC uses only self-generated code snippets and tests to construct reliable preference pairs.
  arXiv  Detail & Related papers  (2024-11-20T02:03:16Z)
• Reference Trustable Decoding: A Training-Free Augmentation Paradigm for Large Language Models [79.41139393080736]
  Large language models (LLMs) have rapidly advanced and demonstrated impressive capabilities. In-Context Learning (ICL) and. Efficient Fine-Tuning (PEFT) are currently two mainstream methods for augmenting. LLMs to downstream tasks. We propose Reference Trustable Decoding (RTD), a paradigm that allows models to quickly adapt to new tasks without fine-tuning.
  arXiv  Detail & Related papers  (2024-09-30T10:48:20Z)
• Adversarial Robustification via Text-to-Image Diffusion Models [56.37291240867549]
  Adrial robustness has been conventionally believed as a challenging property to encode for neural networks. We develop a scalable and model-agnostic solution to achieve adversarial robustness without using any data.
  arXiv  Detail & Related papers  (2024-07-26T10:49:14Z)
• Code Representation Learning At Scale [75.04686476303436]
  We fuel code representation learning with a vast amount of code data via a two-stage pretraining scheme. We first train the encoders via a mix that leverages both randomness in masking language modeling and the structure aspect of programming language. We then enhance the representations via contrastive learning with hard negative and hard positive constructed in an unsupervised manner.
  arXiv  Detail & Related papers  (2024-02-02T22:19:15Z)
• Finetuning Large Language Models for Vulnerability Detection [0.0]
  This paper presents the results of finetuning large language models (LLMs) for the task of detecting vulnerabilities in source code. We leverage WizardCoder, a recent improvement of the state-of-the-art LLM StarCoder, and adapt it for vulnerability detection through further finetuning. The key contributions are finetuning the state-of-the-art code LLM, WizardCoder, increasing its training speed without the performance harm, optimizing the training procedure and regimes, handling class imbalance, and improving performance on difficult vulnerability detection datasets.
  arXiv  Detail & Related papers  (2024-01-30T13:46:49Z)
• InferAligner: Inference-Time Alignment for Harmlessness through Cross-Model Guidance [56.184255657175335]
  We develop textbfInferAligner, a novel inference-time alignment method that utilizes cross-model guidance for harmlessness alignment. Experimental results show that our method can be very effectively applied to domain-specific models in finance, medicine, and mathematics. It significantly diminishes the Attack Success Rate (ASR) of both harmful instructions and jailbreak attacks, while maintaining almost unchanged performance in downstream tasks.
  arXiv  Detail & Related papers  (2024-01-20T10:41:03Z)
• MBR and QE Finetuning: Training-time Distillation of the Best and Most Expensive Decoding Methods [13.56549575939123]
  We propose finetuning and QE finetuning to mitigate the model-perplexity-vs-quality mismatch. We show that even with self-training, these finetuning methods significantly outperform the base model. These findings suggest new ways to leverage monolingual data to achieve improvements in model quality that are on par with, or even exceed, improvements from human-curated data.
  arXiv  Detail & Related papers  (2023-09-19T23:39:07Z)
• The Wisdom of Hindsight Makes Language Models Better Instruction Followers [84.9120606803906]
  Reinforcement learning has seen wide success in finetuning large language models to better align with instructions via human feedback. In this paper, we consider an alternative approach: converting feedback to instruction by relabeling the original one and training the model for better alignment in a supervised manner. We propose Hindsight Instruction Relabeling (HIR), a novel algorithm for aligning language models with instructions.
  arXiv  Detail & Related papers  (2023-02-10T12:16:38Z)
• Differentiable Prompt Makes Pre-trained Language Models Better Few-shot Learners [23.150999852147283]
  This study proposes a novel pluggable, and efficient approach named DifferentiAble pRompT (DART) It can convert small language models into better few-shot learners without any prompt engineering. A comprehensive evaluation of standard NLP tasks demonstrates that the proposed approach achieves a better few-shot performance.
  arXiv  Detail & Related papers  (2021-08-30T12:29:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.