English(EN) Near-Duplicate Image Selection Transfers Preferences in Vision-Language Models

AI模型通过精心策划的近乎重复图像转移偏好

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-21 04:00

研究人员展示了一种称为潜意识学习的现象，即一个AI模型可以通过数据策划将偏好转移给另一个模型，即使源数据保持不变。通过根据标题可能性选择近乎重复的图像，策展人影响了下游模型表现出相似的偏好。在LLaVA-1.5、Qwen2.5-VL、Qwen3-VL和Gemma 4等模型中观察到了这种效应，并且所选的偏好在模型代际之间转移。 AI

影响展示了一种微妙形式的AI对齐风险，即模型偏好可以通过数据策划转移，从而影响未来的模型训练。

排序理由该集群基于一篇详细介绍AI模型行为新发现的学术论文。[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: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群基于一篇详细介绍AI模型行为新发现的学术论文。[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, 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: 4 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Jaeyoung Kim, Eunseok Kim, Dongsuk Jang, Seunghyun Kim · 2026-10-21 04:00

近乎重复的图像选择迁移了视觉语言模型的偏好

arXiv:2610.27943v1 Announce Type: new Abstract: Subliminal learning is the phenomenon in which a model transmits its preferences to another model through data that do not explicitly express them. As models increasingly curate training data for other models, we ask whether this tr…

本文如何被排名

报道来源 [1]

近乎重复的图像选择迁移了视觉语言模型的偏好

相关实体

相关话题