Schema-Adaptive Action-Conditioned JEPA for Cross-Machine CNC Transfer under Partial Sensor Overlap

arXiv cs.LG Papers
jepa action-conditioned cross-machine-transfer cnc sensor-overlap industrial-ai transfer-learning

Summary

The paper presents SAAC-JEPA, a schema-adaptive action-conditioned JEPA model for transferring predictive representations across CNC machines with partial sensor overlap, demonstrating that cross-machine adaptation requires distinct evaluation beyond source-domain accuracy.

arXiv:2609.16071v1 Announce Type: new Abstract: Cross-machine deployment of industrial world models requires transfer across changes in dynamics, sensing interfaces, sampling regimes, and control units. We study a schema-adaptive action-conditioned Joint-Embedding Predictive Architecture (SAAC-JEPA) for CNC dynamics, where the source machine has 17 canonical sensor channels and the target shares only 10. Evaluation uses group-disjoint source splits, source-only normalization, held-out self-supervised validation, unit audits, and a sealed target test after model locking. Across five seeds, JEPA pretraining gives no clean-source forecasting gain: scratch and pretrained-body models obtain \(\mathrm{RMSE}=0.811\pm0.022\) and \(0.813\pm0.022\). A source-only search over 20 candidates selects a schema-consistent action-conditioned JEPA after seven-seed stability checks. On the confirmatory target pass, the locked model reaches zero-shot \(\mathrm{RMSE}=0.546\), \(R^2=0.012\), and \(\mathrm{NLL}=0.52\), outperforming persistence but not RevIN-equipped PatchTST and iTransformer baselines (\(0.503\) and \(0.498\)). A pre-declared paired ablation shows that RevIN in the same architecture improves RMSE to \(0.495\pm0.004\) over three seeds, but degrades target calibration (\(\mathrm{NLL}=20.6\)) on stationary context windows. A pre-lock adaptation sweep further reduces RMSE to \(0.520\) with limited target support. These results show that source-domain forecasting accuracy alone is insufficient to assess industrial predictive representations, and that cross-machine adaptation under partial sensor overlap is a distinct evaluation axis.
Original Article
View Cached Full Text

Cached at: 09/16/26, 08:39 AM 

# Schema-Adaptive Action-Conditioned JEPA for Cross-Machine CNC Transfer under Partial Sensor Overlap
Source: [https://arxiv.org/abs/2609.16071](https://arxiv.org/abs/2609.16071)
[View PDF](https://arxiv.org/pdf/2609.16071)

> Abstract:Cross\-machine deployment of industrial world models requires transfer across changes in dynamics, sensing interfaces, sampling regimes, and control units\. We study a schema\-adaptive action\-conditioned Joint\-Embedding Predictive Architecture \(SAAC\-JEPA\) for CNC dynamics, where the source machine has 17 canonical sensor channels and the target shares only 10\. Evaluation uses group\-disjoint source splits, source\-only normalization, held\-out self\-supervised validation, unit audits, and a sealed target test after model locking\. Across five seeds, JEPA pretraining gives no clean\-source forecasting gain: scratch and pretrained\-body models obtain \\\(\\mathrm\{RMSE\}=0\.811\\pm0\.022\\\) and \\\(0\.813\\pm0\.022\\\)\. A source\-only search over 20 candidates selects a schema\-consistent action\-conditioned JEPA after seven\-seed stability checks\. On the confirmatory target pass, the locked model reaches zero\-shot \\\(\\mathrm\{RMSE\}=0\.546\\\), \\\(R^2=0\.012\\\), and \\\(\\mathrm\{NLL\}=0\.52\\\), outperforming persistence but not RevIN\-equipped PatchTST and iTransformer baselines \(\\\(0\.503\\\) and \\\(0\.498\\\)\)\. A pre\-declared paired ablation shows that RevIN in the same architecture improves RMSE to \\\(0\.495\\pm0\.004\\\) over three seeds, but degrades target calibration \(\\\(\\mathrm\{NLL\}=20\.6\\\)\) on stationary context windows\. A pre\-lock adaptation sweep further reduces RMSE to \\\(0\.520\\\) with limited target support\. These results show that source\-domain forecasting accuracy alone is insufficient to assess industrial predictive representations, and that cross\-machine adaptation under partial sensor overlap is a distinct evaluation axis\.

## Submission history

From: Ayoub Louaye Bouaziz \[[view email](https://arxiv.org/show-email/17640a10/2609.16071)\] **\[v1\]**Sun, 13 Sep 2026 15:10:06 UTC \(582 KB\)

Similar Articles

Diagnosing JEPA World Models with Action-Conditioned Predictive Consistency

arXiv cs.LG

Introduces Action-Conditioned Predictive Consistency (ACPC), a diagnostic for JEPA world models that measures how clean and perturbed observations diverge under action-conditioned rollouts, with theoretical bounds on prediction error and planner cost. Experiments on visual control tasks validate the diagnostic across models like LeWM and PLDM.

One Future, Every Robot: Label-Efficient Collective-State Prediction with Decentralized JEPA

Hugging Face Daily Papers

This paper introduces Collective-State JEPA (CS-JEPA), a recurrent joint-embedding predictive architecture that lets every robot in a swarm predict the same future collective state from local observations and limited messages. The method shows label-efficient improvements in prediction error and inter-robot agreement, plus planning-relevant value estimation.

JEPA-Anything: Learning Predictive Models across Different Worlds

Hugging Face Daily Papers

JEPA-Anything presents a domain-agnostic framework based on orthogonal predictive factorization for learning predictive models across diverse systems like vision, biology, and control, with demonstrated improvements and experimental validation.