English(EN) Over-Personalization Is a Decision Failure: Generation-Induced Apply Bias in LLMs

新的ABIDE方法揭示LLM过度个性化偏差

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-01 04:00

研究人员发现大型语言模型（LLM）中存在一种称为“过度个性化”的现象，即模型在应抑制的上下文中错误地应用存储的偏好。开发了一种名为ABIDE（Apply-Bias Investigation via Decision-score）的新方法来分析此问题。ABIDE揭示，这种失败源于“生成引起的应用偏差”，即LLM生成答案的目标会将其决策转向应用偏好，即使敏感性在很大程度上得以维持。研究人员证明，通过在解码过程中减去一个偏差标量，可以减少不正确偏好的泄露，同时仍能满足模型的预期响应。 AI

影响这项研究通过解决一个基本的决策失败问题，有望实现更可靠、偏差更小的个性化LLM输出。

排序理由该集群包含一篇详细介绍LLM行为的新方法和发现的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 25 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍LLM行为的新方法和发现的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Haeun Jang, Yonghyun Jun, Hwanhee Lee · 2026-10-01 04:00

过度个性化是决策失误：LLM 中由生成引起的应用偏差

arXiv:2609.34284v2 Announce Type: replace Abstract: Personalized LLMs must decide, for each stored preference, whether the current context calls for applying or suppressing it, which we call its applicability. They frequently over-personalize, applying preferences the context rul…

本文如何被排名

报道来源 [1]

过度个性化是决策失误：LLM 中由生成引起的应用偏差

相关实体

相关话题