Hidden in the Request: Explaining Unethical LLM Compliance through Token Relevance
Although Large Language Models (LLMs) are aligned to optimize for both helpfulness and harmlessness, these dual objectives may conflict, inevitably leading to alignment failures. This work systematically investigates instances where LLMs fail to exhibit ethical behavior. To understand the underlying mechanics of these vulnerabilities, we introduce a probing methodology that presents unethical scenarios to LLMs in three distinct structural modalities: objective classification tasks, subjective first-person statements, and direct requests for assistance. We find that model performance degrades in the request-for-assistance-based form. Using Layer-wise Relevance Propagation (LRP), we trace this discrepancy to an attribution bias: the model places greater emphasis on benign task-framing tokens (e.g., "Can you help me...") than on tokens signaling the underlying unethical behavior (e.g., "without getting caught"), which we term cue-tokens. We hypothesize that this under-attribution contributes to harmful compliance. To test this, we introduce two LRP-guided decoding methods that steer generation toward trajectories more relevant to cue tokens. Empirical evaluations show that these interventions promote safer responses, supporting cue-token attribution's role in compliance failures.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
FinSafetyBench: Evaluating LLM Safety in Real-World Financial Scenarios
Large language models (LLMs) are increasingly applied in financial scenarios. However, they may produce harmful outputs, including facilitating illegal activities or unethical behavior, posing serious compliance risks. T…
The Rogue Scalpel: Activation Steering Compromises LLM Safety
Activation steering is a promising technique for controlling LLM behavior by adding semantically meaningful vectors directly into a model's hidden states during inference. It is often framed as a precise, interpretable, …
Unintended Misalignment from Agentic Fine-Tuning: Risks and Mitigation
Beyond simple text generation, Large Language Models (LLMs) have evolved into agentic systems capable of planning and interacting with external tools to solve complex tasks. This evolution involves fine-tuning LLMs on ag…
Code GenerationText GenerationThe Art of Saying No: Contextual Noncompliance in Language Models
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 noncomplian…
The Safety Relay in Roleplay Jailbreaks: A Component-Resolved Causal Analysis of Harm Recognition and Refusal
Large language models are trained to follow instructions while refusing harmful requests. Jailbreaks exploit this balance to elicit content a model would ordinarily reject. Roleplay jailbreaks are especially concerning: …