Reasoning Denoiser: Denoising Reasoning Traces for Hallucination Detection in Large Reasoning Models

  • 类型:arxiv
  • 标识:2607.22098
  • 链接:https://arxiv.org/abs/2607.22098
  • 主分类:rag
  • 形态:method
  • 被引:0
  • 被引来源:Semantic Scholar + OpenAlex
  • S2被引:0
  • OpenAlex被引:0
  • 影响力被引:0
  • TLDR:RedE leverages final-answer attention as an automatic supervision signal to shape the step-level representation space, yielding refined embeddings in which noisy steps can be reliably identified and filtered and improves detection performance over competitive baselines.
  • OpenAlex ID:W7171394929
  • OpenAlex DOI:10.48550/arxiv.2607.22098
  • DOI:10.48550/arxiv.2607.22098
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://doi.org/10.48550/arxiv.2607.22098
  • OpenAlex更新:2026-08-25
  • 待LLM分类:否
  • 标题中文:Reasoning Denoiser:用于大推理模型幻觉检测的推理轨迹去噪
  • TLDR中文:RedE 利用 final-answer attention 作为自动监督信号来塑造 step-level 表征空间,使其中的噪声步骤可被可靠识别与过滤,并在检测性能上超越有竞争力的基线。
  • 来源文件
  • /inbox/tom/_candidates/2026-07-28-agent-rag-longcontext-candidates.json
  • /inbox/tom/_candidates/2026-07-29-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]