MI-MIDI：基于探测、透镜与调控的文本生成MIDI模型的可机理解释
MI-MIDI: Mechanistic Interpretability of Text-to-MIDI Generation Models via Probing, Lenses and Steering
AI总结：
本研究针对两款文本生成MIDI模型，采用探测、透镜等方法开展可机理解释，揭示架构对音乐结构的塑造作用，提出双取向调控协议并提供音乐概念追踪工具包。
AI中文摘要：
音乐生成的可机理解释研究多聚焦于音频模型，符号模型则未被充分探索。我们分析了两款设计迥异的公开文本生成MIDI系统：专用的编码器-解码器模型text2midi，以及通过线性探测、logit透镜、tuned透镜、激活修补与均值差调控扩展了MIDI token的Llama~3.2~1B模型MIDI-LLM。通过这些方法，我们还原了具有音乐意义的结构，并揭示架构如何塑造该结构及其可控制性。音高、乐器、和声与织体在两款模型中均可线性解码；text2midi会在各层逐步优化预测，而MIDI-LLM则主要在继承的文本基础上运作，直至后期才急剧转向音乐词汇；激活修补识别出提示驱动的乐器迁移在后期的相应衰减。调控在两款系统中均会产生音域与复调的双向变化，在MIDI-LLM中还会产生速度与能量的变化。我们的双取向协议分离了定向控制，显示全层干预在text2midi中具有鲁棒性，但在MIDI-LLM中会累积破坏性影响。这些结果共同为追踪与控制符号生成器中的音乐概念提供了实用工具包，音频示例可在演示网站获取。
英文摘要：
Mechanistic interpretability of music generation has concentrated on audio models, leaving symbolic models largely unexplored. We analyze two public text-to-MIDI systems of contrasting design: the purpose-built encoder--decoder text2midi and MIDI-LLM, a Llama~3.2~1B model extended with MIDI tokens using linear probing, the logit and tuned lenses, activation patching and difference-in-means steering. Across these methods, we recover musically meaningful structure and show how architecture shapes its formation and control. Pitch, instrumentation, harmony and texture are linearly decodable in both models. text2midi refines predictions gradually across depth, whereas MIDI-LLM works largely in its inherited textual basis before a sharp late rotation into the musical vocabulary; patching identifies a matching late attenuation of prompt-driven instrument transfer. Steering produces bidirectional changes in register and polyphony in both systems, and in tempo/energy in MIDI-LLM. Our two-orientation protocol isolates directional control and shows that all-layer interventions are robust in text2midi but accumulate disruptively in MIDI-LLM. Together, the results provide a practical toolkit for tracing and controlling musical concepts in symbolic generators. Audio examples are available on a demo website.
