通过激活调控实现音乐属性解耦，无需重训练即可精确控制音高和时长。
Latent Space Disentanglement via Activation Steering for Interpretable Attribute Control in Symbolic Music Generation
- 在推理阶段通过激活向量定向调节，实现对音高和时长的精细控制。
- 线性关系验证显示调控强度与属性变化高度相关，相关系数达0.93以上。
- 采用正交化框架减少多属性干扰，适合需要可解释控制的音乐生成场景。
基于Transformer的架构显著推动了复杂符号序列的生成，但在离散信号属性的细粒度、可解释控制方面仍存在明显差距。本文研究了多轨音乐Transformer（MMT）的机制可解释性，提出一种无需重训练的推理阶段激活调控框架，以填补这一空白。利用均值差（DiffMean）方法，我们在残差流中识别出音高和时长的潜在方向。实验验证了该领域的线性表示假设，调控强度与属性变化之间相关性极高。为解决多属性调控中的特征纠缠问题，引入基于格拉姆-施密特正交化的双路调控框架。结果表明，这种几何解耦有效降低了概念干扰与信号退化，即使在强自回归条件约束下也能实现独立且确定性的属性控制。
原文摘要 · Abstract (English)
Transformer-based architectures have significantly advanced the generation of complex symbolic sequences, yet a significant gap remains in achieving fine-grained, interpretable control over discrete signal attributes. This paper investigates the mechanistic interpretability of the Multitrack Music Transformer (MMT) and proposes a framework for deterministic attribute modulation without retraining to bridge this gap via inference-time activation steering. Utilizing the Difference-in-Means (DiffMean) methodology, we isolate latent directions for signal attributes, specifically Pitch and Duration, within the residual stream. We validate the Linear Representation Hypothesis in this domain, achieving high correlation between steering magnitude and attribute shift. To address the inherent feature entanglement in multi-attribute steering, we introduce a Dual Steering framework utilizing Gram-Schmidt Orthogonalization. Experimental results demonstrate that this geometric decoupling reduces conceptual interference and signal degradation compared to naive vector addition, enabling independent deterministic control even against strong autoregressive conditioning.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。