Words in Motion: Extracting Interpretable Control Vectors for Motion Transformers

TL;DR

This paper introduces a method to interpret and modify hidden states in motion transformers using control vectors, enabling better understanding and zero-shot adaptation with minimal computational cost.

Contribution

We propose a novel approach to extract and refine interpretable control vectors from hidden states of motion transformers for improved interpretability and generalization.

Findings

High probing accuracy indicates latent space regularities.

Control vectors can modify predictions while preserving feasibility.

Refinement with sparse autoencoders improves linearity of modifications.

Abstract

Transformer-based models generate hidden states that are difficult to interpret. In this work, we analyze hidden states and modify them at inference, with a focus on motion forecasting. We use linear probing to analyze whether interpretable features are embedded in hidden states. Our experiments reveal high probing accuracy, indicating latent space regularities with functionally important directions. Building on this, we use the directions between hidden states with opposing features to fit control vectors. At inference, we add our control vectors to hidden states and evaluate their impact on predictions. Remarkably, such modifications preserve the feasibility of predictions. We further refine our control vectors using sparse autoencoders (SAEs). This leads to more linear changes in predictions when scaling control vectors. Our approach enables mechanistic interpretation as well as zero-shot generalization to unseen dataset characteristics with negligible computational overhead.