大型语言模型在明知是虚假陈述的情况下，仍然倾向于相信它们。微调测试表明，这些模型倾向于自信地将错误信息当作事实呈现。这个问题在《纽约时报》记者尝试使用谷歌的 Gemini 聊天机器人卖房时得到了突出体现，这表明了依赖当前 AI 的事实准确性所面临的实际挑战。 AI
影响 模型在有警告的情况下仍然倾向于虚假信息，这可能会限制它们在关键应用中的可靠性。
排序理由 该集群讨论了关于 LLM 在虚假信息方面行为的研究结果。
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
📰 LLMs believe false statements even after explicit warnings that they're false Fine-tuning tests show "bias ... toward confidently representing the claims as true." 📰 Source: Ars Technica 🔗 Link: https://arstechnica.com/ai/2026/05/llms-believe-false-statements-even-after-explici…
🎮 I Love Forza Horizon 6‘s PS1-Looking Train Driver I'll never get tired of low-poly people hanging out in big AAA video games 📰 Source: Kotaku 🔗 Link: https://kotaku.com/i-love-forza-horizon-6s-ps1-looking-train-driver-2000700298 # AI # ArtificialIntelligence
📊 Selling a house with AI Stuart A. Thompson, a New York Times technology journalist, used Google’s Gemini chatbot…Tags: chatbot, house, New York Times, sales, Stuart A. Thompson 📰 Source: FlowingData 🔗 Link: https://flowingdata.com/2026/05/28/selling-a-house-with-ai/ # AI # Arti…