第十一章 · Section 11
Auditory Processing (A) · 听觉处理
Auditory Processing (A): The ability to discriminate, remember, reason, and work creatively on auditory stimuli, which may consist of tones and speech units.
听觉处理（A）：对听觉刺激（可由音调与言语单位构成）进行辨别、记忆、推理与创造性加工的能力。
The ability to hear, blend, and segment phonemes in words.
Sample: "Do 'tref' and 'gref' rhyme?" / "Repeat the following word: [audio]"
清晰听到、混合并切分单词中音素的能力。
样例："'tref' 和 'gref' 押韵吗？" / "重复以下单词：[音频]"
The ability to transcribe a spoken audio signal to text.
Sample: "Transcribe this audio:" / "Transcribe this TED talk:"
把口语音频信号转写为文本的能力。
样例："转写这段音频：" / "转写这场 TED 演讲："
Quality and responsiveness of the AI's synthesized voice.
Sample: "Say this sentence: 'Wait, you mean the tickets were free this whole time?'"
AI 合成语音的质量与响应性。
样例："说出这句话：'等等，你是说这些票一直都是免费的吗？'"
The ability to recognize and maintain a musical beat.
Sample: "Repeat the following rhythm: [audio]" / "Are these two rhythms the same? [audio]"
识别并保持音乐节拍的能力。
样例："重复以下节奏：[音频]" / "这两个节奏相同吗？[音频]"
The ability to judge simple musical patterns.
Sample: "Which note is higher?" / "Identify the musically anomalous part."
判断简单音乐模式的能力。
样例："哪个音符更高？" / "找出音乐中异常的部分。"
评估细节与 AI 表现
Assessment Details & AI Performance
Assessment Details. See Appendix I for further details on how to assess auditory processing capabilities concretely.
AI system performance. The table summarizes current AI system performance on Auditory Processing (A) tasks. GPT-4 had no audio processing capabilities, while GPT-5's capabilities are appreciable but incomplete.
Model | Phonetic (1%) | Speech Recognition (4%) | Voice (3%) | Rhythmic (1%) | Musical (1%) | Total
GPT-4 | 0% | 0% | 0% | 0% | 0% | 0%
GPT-5 | 0% | 4% | 2% | 0% | 0% | 6%
评估细节：如何在具体层面评估听觉处理，参见附录 I。
AI 系统表现：下表汇总当前 AI 系统在听觉处理（A）任务上的表现。GPT-4 完全没有音频处理能力；GPT-5 的能力可观但不完整。
模型 | 语音编码(1%) | 语音识别(4%) | 语音(3%) | 节奏(1%) | 音乐判断(1%) | 合计
GPT-4 | 0% | 0% | 0% | 0% | 0% | 0%
GPT-5 | 0% | 4% | 2% | 0% | 0% | 6%