Benchmarking Behavioral Steerability in Behavior Foundation Models

2026-10-07

AI Summary

This study addresses the critical question of whether action foundation models can reliably generate behaviors aligned with user intent, a problem for which systematic evaluation methods remain lacking. To bridge this gap, this work introduces the novel concept of "action controllability" and establishes the RoboSteer benchmark alongside a three-tier hierarchical evaluation framework. Leveraging a unified multimodal motion corpus, the authors conduct a large-scale empirical study encompassing nine mainstream models. This research delivers the first systematic cross-model analysis of action controllability, providing a crucial evaluation paradigm for precise intent alignment and reliable behavior generation in general-purpose embodied artificial intelligence.

Abstract

Behavior Foundation Models (BFMs) are emerging as a paradigm for translating human intentions into executable humanoid behaviors. As these models evolve beyond behavior generation toward general-purpose behavioral systems, a fundamental question arises: can they be reliably steered according to user intentions? In this paper, we introduce the concept of behavioral steerability, defined as the ability of BFMs to faithfully generate behaviors that satisfy user-specified intentions. To study this capability, we present RoboSteer, the first benchmark for behavioral steerability in BFMs. RoboSteer organizes behavioral steerability into a three-level hierarchy-Conditional Steering, Constraint Steering, and Compositional Steering-and establishes a unified evaluation framework supported by a large-scale multimodal motion corpus. Using RoboSteer, we conduct the first large-scale empirical study of behavioral steerability across 9 existing BFMs. We view behavioral steerability as more than a capability for controlling motion: it concerns how embodied systems translate human intentions into purposeful actions. We hope RoboSteer will advance research on intention realization as a foundation for general-purpose embodied intelligence.

Problem

• Behavior Foundation Models
• Behavioral Steerability
• Benchmarking
• Humanoid Behaviors
• Embodied Intelligence

Innovation

• Behavioral Steerability
• Behavior Foundation Models
• RoboSteer Benchmark
• Multimodal Motion Corpus
• Embodied Intelligence