1
直接问你要钱还是坚持使命，Anthropic 面试流程曝光
Anthropic面试竟直问“股价归零你怕不怕”，使命与金钱的灵魂拷问曝光
IT之家 8 月 24 日消息，据外媒 Axios 今日报道，熟悉 Anthropic 招聘流程的人士透露，这家前沿人工智能实验室会在招人时问一个相当直接的问题： 如果公司的使命与金钱发生冲突时 ， 你会如何选择 ？ 据报道，所有 Anthropic 求职者都需要参与一次文化面试，让公司评估应聘者的…
2
Which Values Do LLMs Confuse? A Schwartz-Based Recognition Study
研究揭开了LLM在识别人类价值观时的困惑点，基于Schwartz理论的首个系统评估实验
arXiv:2607.20270v1 Announce Type: new Abstract: Large language models are increasingly evaluated through the values they endorse, but such evaluations…
3
Value Leakage: An LLM's Answers Are Silently Shaped by Its Own Values
当大模型回答问题时，其内置价值观正悄然左右输出，揭示被忽视的“价值泄漏”现象
arXiv:2607.14345v1 Announce Type: new Abstract: People use language models for practical questions whose answers are difficult to verify. We show that…
4
AI puts B Corps' values to the test
AI正在考验B型企业价值观，看科技如何挑战商业向善的边界。
Article URL: https://www.fastcompany.com/91568793/ai-puts-b-corps-values-to-the-test Comments URL: https://news.ycombinator.com/item?id=48782817 Point…
5
Small edits, large models: How Wikipedia advocacy shapes LLM values
维基百科上的小众编辑如何悄然重塑大模型价值观，揭示训练数据背后的隐形影响力
arXiv:2606.24890v1 Announce Type: cross Abstract: Can a small group of volunteers shape how AI systems discuss animal welfare, just by editing Wikiped…
6
AI Pluralism and the Worlds It Misses
探讨AI多元主义主流框架的遗漏之处，质疑“表征多样性”这一常见视角的深层局限。
arXiv:2606.16167v1 Announce Type: new Abstract: AI pluralism is often framed as a problem of representing diverse values, preferences, users, or outpu…
7
LLMs Contain Multitudes: How Deployment Context Reshapes Model-Level Preferences and Values
研究指出LLM的偏好和价值观并非稳定不变，部署上下文会显著重塑模型层面的价值取向，挑战传统评估假设。
arXiv:2606.13944v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly characterised in recent evaluation work as having stable…
8
Position: Align AI to Our Aspirations, Not Our Flaws
AI对齐应引导系统向人类更高追求对齐，而非迎合既有缺陷，这篇论文提出了颠覆性的对齐目标视角。
arXiv:2606.13755v1 Announce Type: cross Abstract: We argue that aligning AI to aggregated human preferences is the wrong target. With current technolo…
9
Dao Heart 3.13 a symbolic safety layer for value drift and AI alignment research
一个符号化的AI安全层，专治价值观漂移和对齐难题，开源研究项目值得关注。
Article URL: https://github.com/Mankirat47/Dao-Heart-3.13 Comments URL: https://news.ycombinator.com/item?id=48442582 Points: 1 # Comments: 0
10
Ask HN: Are we as society going to let LLM companies take all the values?
当LLM巨头疯狂追逐利润，我们是否正在默许它们窃取社会共同价值？这篇HN热帖引发扎心思考。
I am not that young, but I'm not that old. I used to be a child, and thought that the adults already figured things out and I can be at peace. One of …
11
My Latest LLM Workflow and Modern Engineering Values
作者分享结合现代工程价值观的LLM工作流，用几个prompts搞定自定义ROM hack，看点十足的AI编程实战案例。
Article URL: https://cpojer.net/posts/modern-engineering-values Comments URL: https://news.ycombinator.com/item?id=48384354 Points: 2 # Comments: 0
12
Greener Than Humans? Environmental Attitudes in Large Language Models
研究揭示LLM在环境态度上比人类更环保，但存在偏见和价值观差异，引发AI伦理新思考。
arXiv:2606.02741v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used in sustainability-related decision support, reporti…
13
ValueGround: Evaluating Culture-Conditioned Visual Value Grounding in MLLMs
论文提出ValueGround基准，评估多模态大模型对不同文化背景下的视觉价值理解能力，揭示现有模型在文化适应性上的不足。
arXiv:2604.06484v3 Announce Type: replace Abstract: Cultural values are expressed not only through language but also through visual scenes and everyda…
14
Teaching Values to Machines: Simulating Human-Like Behavior in LLMs
心理学理论赋能大模型，诱导LLM展现类人价值行为，探索AI对齐新路径
arXiv:2605.30036v1 Announce Type: new Abstract: Large Language Models (LLMs) demonstrate a remarkable capacity to adopt different personas and roles; …
15
Identifying and Understanding Human Values in Text: A Tailorable LLM-based Architecture
基于LLM的新架构，精准识别文本中的人类价值观，可定制化分析方案
arXiv:2605.27373v1 Announce Type: new Abstract: As intelligent systems become more autonomous, the scientific community focuses on creating decision-m…
16
Cultural Value Alignment Via Latent Activation Steering in Large Language Models
用潜在激活引导破解大模型文化同质化，让AI价值观更贴合人类多样文明
arXiv:2605.26365v1 Announce Type: new Abstract: Large Language Models (LLMs) often exhibit homogenized cultural perspectives. While the World Values S…
17
EvalMORAAL: Interpretable Chain-of-Thought and LLM-as-Judge Evaluation for Moral Alignment in Large Language Models
透明思维链框架 EvalMORAAL，通过双评分法和模型裁判评审评估20个LLM在55国价值观数据上的道德对齐
arXiv:2510.05942v3 Announce Type: replace Abstract: We present EvalMORAAL, a transparent chain-of-thought (CoT) framework that uses two scoring method…
18
Bitwarden 被私募基金收购了吗？
开源密码管理器Bitwarden悄然涨价、价值观调整，背后是资本化转型的信号，老用户很敏感。
省流：目前没有证据表明 Bitwarden 被收购。 今天半夜看到消息：《Bitwarden 已悄然被私募基金收购》： 本着不信的原则，去研究了一番。 目前来看，并没有任何公开信息能证明 Bitwarden 已经被收购。 不过最近这段时间，它确实发生了一些变化。 发生了什么？ 10年来第一次涨价 前
19
AI safety needs social scientists
OpenAI发文论证，长期AI安全研究亟需社会科学家参与，以解决人类心理、偏见与理性不确定性，促进ML与社科跨界协作。
We’ve written a paper arguing that long-term AI safety research needs social scientists to ensure AI alignment algorithms succeed when actual humans a…
20
Humility: An Essential Value
设计路上，谦逊是照亮成长与连接的明灯，一场关于核心价值的内省之旅。
Humility, a designer’s essential value—that has a nice ring to it. What about humility, an office manager’s essential value? Or a dentist’s? Or a libr…