# When Is a Learned Command Adapter Worth It? Closed-Loop Identification and Counterfactual Auditing of Frozen Locomotion Policies

arXiv: https://arxiv.org/abs/2607.21867 · HTML: https://mlsanity.com/paper/2607.21867/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 40 references that are in the map.
- First posted on arXiv 23 Jul 2026.

## Builds on (most cited first, 24 of 24)

- [Dream to Control: Learning Behaviors by Latent Imagination](https://mlsanity.com/paper/1912.01603/) (2019)
- [Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware](https://mlsanity.com/paper/2304.13705/) (2023)
- [Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning](https://mlsanity.com/paper/2108.10470/) (2021)
- [Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models](https://mlsanity.com/paper/1805.12114/) (2018)
- [When to Trust Your Model: Model-Based Policy Optimization](https://mlsanity.com/paper/1906.08253/) (2019)
- [Diversity is All You Need: Learning Skills without a Reward Function](https://mlsanity.com/paper/1802.06070/) (2018)
- [Learning Quadrupedal Locomotion over Challenging Terrain](https://mlsanity.com/paper/2010.11251/) (2020)
- [Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning](https://mlsanity.com/paper/2109.11978/) (2021)
- [RMA: Rapid Motor Adaptation for Legged Robots](https://mlsanity.com/paper/2107.04034/) (2021)
- [The Option-Critic Architecture](https://mlsanity.com/paper/1609.05140/) (2016)
- [Illuminating search spaces by mapping elites](https://mlsanity.com/paper/1504.04909/) (2015)
- [Temporal Difference Learning for Model Predictive Control](https://mlsanity.com/paper/2203.04955/) (2022)
- [Learning robust perceptive locomotion for quadrupedal robots in the wild](https://mlsanity.com/paper/2201.08117/) (2022)
- [Dynamics-Aware Unsupervised Discovery of Skills](https://mlsanity.com/paper/1907.01657/) (2019)
- [Neural Network Dynamics for Model-Based Deep Reinforcement Learning with Model-Free Fine-Tuning](https://mlsanity.com/paper/1708.02596/) (2017)
- [Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments](https://mlsanity.com/paper/2301.04195/) (2023)
- [Residual Reinforcement Learning for Robot Control](https://mlsanity.com/paper/1812.03201/) (2018)
- [Residual Policy Learning](https://mlsanity.com/paper/1812.06298/) (2018)
- [Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior](https://mlsanity.com/paper/2212.03238/) (2022)
- [A predictive safety filter for learning-based control of constrained nonlinear dynamical systems](https://mlsanity.com/paper/1812.05506/) (2018)
- [Information Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving](https://mlsanity.com/paper/1707.02342/) (2017)
- [Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion](https://mlsanity.com/paper/2401.17583/) (2024)
- [RSL-RL: A Learning Library for Robotics Research](https://mlsanity.com/paper/2509.10771/) (2025)
- [A Statistical Test for the Benefits of Personalizing Interventions](https://mlsanity.com/paper/2607.08951/) (2026)

## Built on by (most cited first, 1 of 1)

- [When Does Test-Time Physical Diagnosis Pay? A Frozen Policy Buys Evidence It Never Reads](https://mlsanity.com/paper/2609.22299/) (2026)

## Abstract (arXiv)

Adding a learned adapter to a frozen, command-conditioned locomotion policy is worthwhile only if the interface exposes improvements that are both real and recoverable from deployment-time observations. We introduce an adapter necessity audit that separates global operating-point gain,same-state counterfactual headroom, deployment gain over a cross-fitted fixed action, and state-allocation gain over a frequency-matched randomized policy. Source-cluster learner refits map these quantities and constraint violations to a GO/NO-GO/ABSTAIN decision. Closed-loop command- response identification provides optional decision features. On Go2, an archived scale-prefix diagnostic finds 5.2% same-state headroom but only 0.55% recovered allocation gain. Our confirmatory audit evaluates direct, scale, heading, and yaw interventions on twenty independent clusters for each of three query distributions induced by direct control, VGCC, and MPC, using 200 full learner refits. At 1% deployment and allocation thresholds and a 5% violation tolerance, direct queries return NO-GO, while VGCC and MPC queries ABSTAIN. VGCC has the largest mean deployment gain (1.34%), but its allocation lower bound is 0.09% and its violation upper bound is 6.25%. A deployment-representative twenty-cluster H1 audit also returns NO-GO, whereas a learner-level synthetic control returns GO. The audit therefore tests whether observable signal justifies state-dependent adaptation rather than presuming that an adapter is valuable.