Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Price Is Not Right: Neuro-Symbolic Methods Outperform VLAs on Structured Long-Horizon Manipulation Tasks with Significantly Lower Energy Consumption

Timothy Duggan, Pierrick Lorang, Hong Lu, Matthias Scheutz

TL;DR

A head-to-head empirical comparison between a fine-tuned open-weight VLA model {\pi}0 and a neuro-symbolic architecture that combines PDDL-based symbolic planning with learned low-level control on structured variants of the Towers of Hanoi manipulation task is presented.

Abstract

Vision-Language-Action (VLA) models have recently been proposed as a pathway toward generalist robotic policies capable of interpreting natural language and visual inputs to generate manipulation actions. However, their effectiveness and efficiency on structured, long-horizon manipulation tasks remain unclear. In this work, we present a head-to-head empirical comparison between a fine-tuned open-weight VLA model π0 and a neuro-symbolic architecture that combines PDDL-based symbolic planning with learned low-level control. We evaluate both approaches on structured variants of the Towers of Hanoi manipulation task in simulation while measuring both task performance and energy consumption during training and execution. On the 3-block task, the neuro-symbolic model achieves 95% success compared to 34% for the best-performing VLA. The neuro-symbolic model also generalizes to an unseen 4-block variant (78% success), whereas both VLAs fail to complete the task. During training, VLA fine-tuning consumes nearly two orders of magnitude more energy than the neuro-symbolic approach. These results highlight important trade-offs between end-to-end foundation-model approaches and structured reasoning architectures for long-horizon robotic manipulation, emphasizing the role of explicit symbolic structure in improving reliability, data efficiency, and energy efficiency. Code and models are available at https://price-is-not-right.github.io

The Price Is Not Right: Neuro-Symbolic Methods Outperform VLAs on Structured Long-Horizon Manipulation Tasks with Significantly Lower Energy Consumption

TL;DR

A head-to-head empirical comparison between a fine-tuned open-weight VLA model {\pi}0 and a neuro-symbolic architecture that combines PDDL-based symbolic planning with learned low-level control on structured variants of the Towers of Hanoi manipulation task is presented.

Abstract

Vision-Language-Action (VLA) models have recently been proposed as a pathway toward generalist robotic policies capable of interpreting natural language and visual inputs to generate manipulation actions. However, their effectiveness and efficiency on structured, long-horizon manipulation tasks remain unclear. In this work, we present a head-to-head empirical comparison between a fine-tuned open-weight VLA model π0 and a neuro-symbolic architecture that combines PDDL-based symbolic planning with learned low-level control. We evaluate both approaches on structured variants of the Towers of Hanoi manipulation task in simulation while measuring both task performance and energy consumption during training and execution. On the 3-block task, the neuro-symbolic model achieves 95% success compared to 34% for the best-performing VLA. The neuro-symbolic model also generalizes to an unseen 4-block variant (78% success), whereas both VLAs fail to complete the task. During training, VLA fine-tuning consumes nearly two orders of magnitude more energy than the neuro-symbolic approach. These results highlight important trade-offs between end-to-end foundation-model approaches and structured reasoning architectures for long-horizon robotic manipulation, emphasizing the role of explicit symbolic structure in improving reliability, data efficiency, and energy efficiency. Code and models are available at https://price-is-not-right.github.io
Paper Structure (29 sections, 2 figures, 4 tables)

This paper contains 29 sections, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Vision-Language-Action Models
  4. Neuro-symbolic Approaches for Task and Motion Planning
  5. Preliminaries: Neuro-Symbolic Planning-Diffusion Models
  6. Comparing State-of-the-Art VLA and Neuro-Symbolic Models
  7. Fine-tuning the Vision Language Action Model
  8. Inference with the Vision Language Action Model
  9. Planning
  10. VLA latent space
  11. VLM
  12. Neuro-symbolic Architecture
  13. Evaluation Methodology
  14. Hardware
  15. Simulation and Experimental Environment
  16. ...and 14 more sections

Figures (2)

  • Figure 1: Overview of the experimental comparison between VLA models and the NSM. Both receive identical sensory inputs from simulation. VLAs produce actions conditioned on either a high-level task description or planner guidance, while the NSM plans symbolically and executes via learned policies. Power and energy are monitored during training and inference.
  • Figure 2: Example observations from the dataset. Left: Agent-view RGB image. Right: Wrist-mounted camera RGB image.