Latent Action Learning Requires Supervision in the Presence of Distractors

TL;DR

This paper demonstrates that in the presence of distractors in videos, latent action learning models require supervision to perform effectively, and introduces a simple modification that significantly improves their quality.

Contribution

The paper shows the necessity of supervision in latent action learning with distractors and proposes LAOM, a simple modification that enhances latent action quality.

Findings

LAOM improves latent action quality by 8x

Supervision with 2.5% of data boosts performance by 4.2x

Supervision is critical for latent action learning with distractors

Abstract

Recently, latent action learning, pioneered by Latent Action Policies (LAPO), have shown remarkable pre-training efficiency on observation-only data, offering potential for leveraging vast amounts of video available on the web for embodied AI. However, prior work has focused on distractor-free data, where changes between observations are primarily explained by ground-truth actions. Unfortunately, real-world videos contain action-correlated distractors that may hinder latent action learning. Using Distracting Control Suite (DCS) we empirically investigate the effect of distractors on latent action learning and demonstrate that LAPO struggle in such scenario. We propose LAOM, a simple LAPO modification that improves the quality of latent actions by 8x, as measured by linear probing. Importantly, we show that providing supervision with ground-truth actions, as few as 2.5% of the full dataset, during latent action learning improves downstream performance by 4.2x on average. Our findings suggest that integrating supervision during Latent Action Models (LAM) training is critical in the presence of distractors, challenging the conventional pipeline of first learning LAM and only then decoding from latent to ground-truth actions.