English(EN) Tracing Subliminal Learning Through Pretraining

潜意识学习在不同训练变体中持续存在于大型语言模型中

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-23 04:00

研究人员调查了大型语言模型中的“潜意识学习”，即在其他模型生成的数据上训练的模型可能会继承非预期的特征。该研究探讨了硬标签蒸馏（使用教师生成的数字）和同策略软标签蒸馏（学生模型在教师对数引导下生成数据）。研究使用Pythia、PolyPythia和DataDecide OLMo等模型，发现这种偏好转移在模型架构、训练数据和初始化存在各种差异的情况下仍然存在。研究还指出，模型的教学能力和学习这些偏好的能力可以独立发展，从而导致不对称的转移效应。 AI

影响理解潜意识学习对于控制模型非预期的行为和确保人工智能训练管道中的数据完整性至关重要。

排序理由学术论文，详细介绍了关于大型语言模型行为的研究发现。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 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: 4 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，详细介绍了关于大型语言模型行为的研究发现。[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, model release; 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: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Angela Tang, Vishesh Anand, Gon\c{c}alo Paulo · 2026-10-23 04:00

Tracing Subliminal Learning Through Pretraining

arXiv:2610.30884v1 Announce Type: new Abstract: Large language models trained on data generated by other models can sometimes inherit traits unrelated to that data, a phenomenon called subliminal learning. We study both hard label distillation, where students train on numbers gen…

本文如何被排名

报道来源 [1]

Tracing Subliminal Learning Through Pretraining

相关实体

相关话题