微软画图和照片应用生成的图像嵌入了看不见的水印

Solidot是至顶网的科技资讯网站，主要面对开源自由软件和关心科技资讯读者群，包括众多中国开源软件的开发者，爱好者和布道者。口号是“奇客的知识，重要的东西”。

Auditing the Synthetic Memoir: Measuring Scene-Level Confabulation in LLM-Generated Autobiography Against the Documented Record of the Life It Describes

（翻译）审计合成回忆录：对照其所描述生活的文档记录度量LLM生成自传中的场景级虚构

Abstract page for arXiv paper 2608.23640: Auditing the Synthetic Memoir: Measuring Scene-Level Confabulation in LLM-Generated Autobiography Against the Documented Record of the Life It Describes

Ethical LLM-Assisted Research: A Framework for Responsible Delegation, Verification, and Epistemic Value

（翻译）伦理的LLM辅助研究：负责任委托、验证与认知价值的框架

Abstract page for arXiv paper 2608.23644: Ethical LLM-Assisted Research: A Framework for Responsible Delegation, Verification, and Epistemic Value

Gated Activation Steering for Reducing Sycophancy & Hallucination in Medical Question Answering

（翻译）用于减少医学问答中谄媚与幻觉的门控激活引导

Abstract page for arXiv paper 2608.23666: Gated Activation Steering for Reducing Sycophancy & Hallucination in Medical Question Answering

Claude最强Fable 5.1发布！8项屠榜，最高降价45%，反蒸馏机制上线

无论你此前让Claude处理什么任务，Fable 5.1都能做更多，并且把最难的部分完成得更出色

LLM-Driven Autonomous Vehicles Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From A New Benchmark

（翻译）LLM驱动的自动驾驶汽车在行人让行中继承人类驾驶员的偏见：来自新基准的结果与启示

Abstract page for arXiv paper 2609.00192: LLM-Driven Autonomous Vehicles Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From A New Benchmark

Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation

（翻译）陷入故事：多轮LLM对话中的叙事俘获

Abstract page for arXiv paper 2609.03407: Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation

ContractEval: Query-Conditioned Execution Matching for Procedural Instruction Conformance

（翻译）ContractEval：面向程序性指令合规的查询条件化执行匹配

Abstract page for arXiv paper 2609.09458: ContractEval: Query-Conditioned Execution Matching for Procedural Instruction Conformance