Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models

  • 类型：arxiv
  • 标识：2609.04714
  • 链接：https://arxiv.org/abs/2609.04714
  • 主分类：risk
  • 形态：method
  • 被引：2
  • 被引来源：Semantic Scholar
  • S2被引：2
  • OpenAlex被引：0
  • 影响力被引：0
  • TLDR：This paper decomposes a response in the safety-tuning dataset into two distinct components: a boilerplate refusal statement and a rationale explaining the refusal, and shows that refusal statements impede accurate discrimination between harmful and benign queries by inducing reliance on superficial cues.
  • OpenAlex ID：W7211887640
  • OpenAlex DOI：10.48550/arxiv.2609.04714
  • DOI：10.48550/arxiv.2609.04714
  • DOI来源：OpenAlex
  • 开放获取：green
  • 开放获取链接：https://doi.org/10.48550/arxiv.2609.04714
  • OpenAlex更新：2026-09-09
  • 待LLM分类：否
  • 标题中文：Refuse without Refusal：面向语言模型安全调优响应的结构性分析以减少误拒
  • TLDR中文：本文将安全微调数据集中的回复拆分为两个独立部分：模板化的拒答声明与解释拒答的理由，并表明拒答声明会诱导模型依赖表层线索，从而妨碍对有害与良性查询的准确区分。
  • 来源文件：
  • /inbox/tom/_candidates/2026-09-08-rag-retrieval-reranking-candidates.json
  • /inbox/tom/_candidates/2026-09-08-agent-rag-longcontext-candidates.json
  • [OpenAlex backfill]
  • [S2 enrich]