Self and Cross-Model Distillation for LLMs: Effective Methods for Refusal Pattern Alignment

• URL: http://arxiv.org/abs/2406.11285v2
• Date: Mon, 02 Dec 2024 05:22:01 GMT
• Title: Self and Cross-Model Distillation for LLMs: Effective Methods for Refusal Pattern Alignment
• Authors: Jie Li, Yi Liu, Chongyang Liu, Xiaoning Ren, Ling Shi, Weisong Sun, Yinxing Xue,
• Abstract summary: Large Language Models (LLMs) like OpenAI's GPT series, Anthropic's Claude, and Meta's LLaMa have shown remarkable capabilities in text generation. Their susceptibility to toxic prompts presents significant security challenges. This paper investigates alignment techniques, including Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF)
• Score: 11.623119255726698
• License:
• Abstract: Large Language Models (LLMs) like OpenAI's GPT series, Anthropic's Claude, and Meta's LLaMa have shown remarkable capabilities in text generation. However, their susceptibility to toxic prompts presents significant security challenges. This paper investigates alignment techniques, including Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF), to mitigate these risks. We conduct an empirical study on refusal patterns across nine LLMs, revealing that models with uniform refusal patterns, such as Claude3, exhibit higher security. Based on these findings, we propose self-distilling and cross-model distilling methods to enhance LLM security. Our results show that these methods significantly improve refusal rates and reduce unsafe content, with cross-model distilling achieving refusal rates close to Claude3's 94.51%. These findings underscore the potential of distillation-based alignment in securing LLMs against toxic prompts.

Related papers

• Quantification of Large Language Model Distillation [22.680566179355335]
  We propose a framework to evaluate and quantify model distillation. Our method addresses two key aspects: (1) Identifying identity cognition contradictions to assess discrepancies in how models perceive and represent identity-related information, and (2) Analyzing multi-granularity response similarities across models to measure the extent of homogenization.
  arXiv  Detail & Related papers  (2025-01-22T03:57:52Z)
• On Evaluating the Durability of Safeguards for Open-Weight LLMs [80.36750298080275]
  We discuss whether technical safeguards can impede the misuse of large language models (LLMs) We show that even evaluating these defenses is exceedingly difficult and can easily mislead audiences into thinking that safeguards are more durable than they really are. We suggest future research carefully cabin claims to more constrained, well-defined, and rigorously examined threat models.
  arXiv  Detail & Related papers  (2024-12-10T01:30:32Z)
• Diversity Helps Jailbreak Large Language Models [16.34618038553998]
  We have uncovered a powerful jailbreak technique that leverages large language models' ability to diverge from prior context. By simply instructing the LLM to deviate and obfuscate previous attacks, our method dramatically outperforms existing approaches. This revelation exposes a critical flaw in current LLM safety training, suggesting that existing methods may merely mask vulnerabilities rather than eliminate them.
  arXiv  Detail & Related papers  (2024-11-06T19:39:48Z)
• Precision Knowledge Editing: Enhancing Safety in Large Language Models [4.241100280846233]
  This work introduces Precision Knowledge Editing (PKE), an advanced technique that builds upon existing knowledge editing methods. PKE achieves finer granularity in toxic content management compared to previous methods like Detoxifying Instance Neuron Modification (DINM) Our experiments demonstrate that PKE significantly reduces the attack success rate (ASR) across various models.
  arXiv  Detail & Related papers  (2024-10-02T23:15:53Z)
• Characterizing and Evaluating the Reliability of LLMs against Jailbreak Attacks [23.782566331783134]
  We focus on 10 cutting-edge jailbreak strategies across three categories, 1525 questions from 61 specific harmful categories, and 13 popular LLMs. We adopt multi-dimensional metrics such as Attack Success Rate (ASR), Toxicity Score, Fluency, Token Length, and Grammatical Errors to thoroughly assess the LLMs' outputs under jailbreak. We explore the relationships among the models, attack strategies, and types of harmful content, as well as the correlations between the evaluation metrics, which proves the validity of our multifaceted evaluation framework.
  arXiv  Detail & Related papers  (2024-08-18T01:58:03Z)
• Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training [67.30423823744506]
  This study addresses a critical gap in safety tuning practices for Large Language Models (LLMs) We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position. DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful
  arXiv  Detail & Related papers  (2024-07-12T09:36:33Z)
• Model Surgery: Modulating LLM's Behavior Via Simple Parameter Editing [63.20133320524577]
  We show that editing a small subset of parameters can effectively modulate specific behaviors of large language models (LLMs) Our approach achieves reductions of up to 90.0% in toxicity on the RealToxicityPrompts dataset and 49.2% on ToxiGen.
  arXiv  Detail & Related papers  (2024-07-11T17:52:03Z)
• Unveiling the Misuse Potential of Base Large Language Models via In-Context Learning [61.2224355547598]
  Open-sourcing of large language models (LLMs) accelerates application development, innovation, and scientific progress. Our investigation exposes a critical oversight in this belief. By deploying carefully designed demonstrations, our research demonstrates that base LLMs could effectively interpret and execute malicious instructions.
  arXiv  Detail & Related papers  (2024-04-16T13:22:54Z)
• Unveiling the Implicit Toxicity in Large Language Models [77.90933074675543]
  The open-endedness of large language models (LLMs) combined with their impressive capabilities may lead to new safety issues when being exploited for malicious use. We show that LLMs can generate diverse implicit toxic outputs that are exceptionally difficult to detect via simply zero-shot prompting. We propose a reinforcement learning (RL) based attacking method to further induce the implicit toxicity in LLMs.
  arXiv  Detail & Related papers  (2023-11-29T06:42:36Z)
• Mind's Mirror: Distilling Self-Evaluation Capability and Comprehensive Thinking from Large Language Models [20.28989820878285]
  Large language models (LLMs) have achieved remarkable advancements in natural language processing. The massive scale and computational demands of these models present formidable challenges when considering their practical deployment in resource-constrained environments.
  arXiv  Detail & Related papers  (2023-11-15T18:56:23Z)
• Gaining Wisdom from Setbacks: Aligning Large Language Models via Mistake Analysis [127.85293480405082]
  The rapid development of large language models (LLMs) has not only provided numerous opportunities but also presented significant challenges. Existing alignment methods usually direct LLMs toward the favorable outcomes by utilizing human-annotated, flawless instruction-response pairs. This study proposes a novel alignment technique based on mistake analysis, which deliberately exposes LLMs to erroneous content to learn the reasons for mistakes and how to avoid them.
  arXiv  Detail & Related papers  (2023-10-16T14:59:10Z)

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.