ARTICLE · 1014071
【高中英语这样学】【'榨干'试题】2026北京昌平高三一模阅读C篇|Part 3: 阅读表达
【高中英语这样学】【'榨干'试题】2026北京昌平高三一模阅读C篇|Part 3: 阅读表达阅读短文，根据短文内容回答问题。 Today's digital world brings a challenge: distinguishing between human‑made and AI‑generated content. For example, AI can create fake images or spread false information, which causes serious concerns. To address this, researchers are exploring a new technique called AI watermarking. AI watermarking is the process of embedding a recognizable, unique signal into the output of an AI model, such as text or an image, to identify that content as AI‑generated. Ideally, an AI watermark should be invisible to the human eye, but easily found by specialized software or computer algorithms. Effective AI watermarking should avoid reducing model performance, resist efforts to forge, remove or change it, and work well with various model architectures. This technique is created during model training by teaching the model to embed a specific signal or identifier in generated content. This process usually involves making small changes to the model during the training stage. After model training and use, specialized algorithms detect the presence of the watermark embedded earlier, thereby checking whether a piece of media was generated by AI. Similar to a physical watermark on paper currency, AI watermarks serve as digital signatures that can demonstrate the origin of a piece of media. Social media networks and news organizations could use them to indicate AI‑created content to readers. Unfortunately, current AI watermarking techniques are unreliable and relatively easy to bypass. Developing long‑lasting AI watermarks that not even determined attackers can remove remains an open research problem. One significant issue is that watermarks are often easy to remove, particularly in text. There’s also the problem of "false positives", where human‑made content is wrongly identified as AI‑generated. Finding the right balance is key. As AI technology advances, improving watermarking techniques remains an important task. 根据短文内容，回答下列问题。 36. What is the function of AI watermarking? (no more than 12 words) 37. Where is an AI watermark embedded? (no more than 10 words) 38. What is one major shortcoming of present AI watermarking techniques? 39. What does “false positives” in the last paragraph refer to? 40. What is the main idea of the passage? 36.To identify AI‑generated content. 解析：细节题，第二段第一句，AI水印用来识别AI生成的内容。 37.Into the output of an AI model. 解析：细节定位第二段，水印嵌入AI模型输出结果中。 38.They are unreliable and easy to bypass / remove. 解析：最后一段首句，现有技术不可靠，容易被绕过清除。 39.Human‑made content is wrongly regarded as AI‑generated. 解析：词义猜测，原文解释false positives：人类创作内容被错误识别为AI生成。 40.It introduces AI watermarking, including its principle, uses and existing technical problems. 解析：主旨概括，文章介绍AI水印：工作原理、用途以及现存技术缺陷。
母题：【高中英语这样学】【‘榨干’试题】2026北京昌平高三一模阅读C篇｜Part 1: 真题演练
练一练｜阅读表达
阅读表达参考答案&解析
声明：部分内容借助AI生成