The Robot Is Not Its Description: GaugeBench for Representation Robustness in Morphology-Aware Policies

2026-10-05

AI Summary

This study addresses the performance collapse of morphology-aware policies under altered robotic description conventions, such as axis orientations and zero positions, noting that prior work fails to disentangle mechanism robustness from representation robustness. To this end, we introduce GaugeBench, a benchmark employing physically equivalent description rewrites to decouple cross-embodiment evaluation into these two dimensions for the first time, validated through interface consistency verification and multi-framework comparative experiments. Our findings reveal that modifying description conventions is more detrimental than changing robot embodiments, with axis inversion identified as the primary cause. We further propose dual-description transfer and cross-convention training strategies that restore retained returns from 3.6% to 80.6%, effectively repairing representation robustness.

Abstract

A robot description does more than specify a physical mechanism: it also encodes arbitrary conventions, such as joint-axis direction, joint-angle zero, and the order and names of links and joints. Morphology-aware policies consume interfaces built from these descriptions, yet cross-embodiment evaluation typically changes the robot while keeping those conventions fixed. This leaves a simple question unanswered: does behavior survive when the robot stays fixed but its description changes? GaugeBench isolates this case by rewriting a fixed mechanism under physically equivalent conventions, verifying that its physics and policy interface are preserved, and then evaluating the same policy weights. The result is stark: three MetaMorph policies score 4030.6 on 80 familiar robots, but only 51.6 when those same robots are equivalently re-described, while 98 genuinely held-out robots score 1489.6. A new description can therefore be more damaging than a new robot. Tracing the failure reveals that axis reversal alone reproduces the collapse, joint-angle zero changes are nearly harmless, and reordering lies between them; moreover, changing joint-state and torque coordinates alone is sufficient to cause the failure, while changing description-derived features alone is not. The same phenomenon appears in ModuMorph and an unrelated PyBullet framework. Yet it is not irreversible: exact two-description transport restores the original controller, and training across equivalent axis conventions raises retained return under axis reversal from 3.6% to 80.6%. Together, these results separate mechanism robustness from representation robustness and show that cross-embodiment evaluation should test both.

Problem

• representation robustness
• morphology-aware policies
• cross-embodiment evaluation
• robot description
• gauge invariance

Innovation

• GaugeBench
• Representation Robustness
• Morphology-Aware Policies
• Cross-Embodiment Evaluation
• Gauge Equivalence