SeT-Diff: Towards Semantic Foundation Models for HPC Telemetry and Time-Series
Quick Answer
SeT-Diff is a novel foundational model for HPC telemetry and time-series, achieving a Mean Absolute Error (MAE) of 0.0470 on reconstruction tasks.
Quick Take
It utilizes a diffusion-based approach to decouple system dynamics from dataset structure, ensuring zero-shot permutation stability and effective data imputation, forecasting, and virtual sensing with a 0.033 MAE in thermal inference.
Key Points
- SeT-Diff is the first foundational model specifically for compute node telemetry.
- Achieved a Mean Absolute Error (MAE) of 0.0470 on reconstruction tasks.
- Exhibits zero-shot permutation stability, maintaining accuracy with sensor shuffling.
- Single pre-trained model performs data imputation, forecasting, and virtual sensing.
- Achieved a 0.033 MAE in thermal inference, enhancing HPC system modeling.
DeepSignal Analysis
What happened
SeT-Diff is introduced as a foundational model for high-performance computing (HPC) telemetry and time-series analysis. It achieves a Mean Absolute Error (MAE) of 0.0470 in reconstruction tasks and demonstrates zero-shot permutation stability, allowing it to maintain accuracy even when sensor data is shuffled.
Key evidence
- SeT-Diff utilizes a diffusion-based approach that conditions the generative process on each sensor's semantic description, decoupling system dynamics from dataset structure.
- Experiments on a real-world supercomputer dataset show SeT-Diff achieving a 0.0470 MAE on reconstruction tasks and a 0.033 MAE in thermal inference.
- The model effectively performs data imputation, forecasting, and virtual sensing, making it a versatile tool for HPC systems.
Why it matters
The development of SeT-Diff addresses the limitations of existing machine learning models in HPC, which often rely on static sensor variables. By providing a flexible and accurate digital twin, it can adapt to changing workloads and sensor metrics, potentially improving the efficiency and reliability of HPC systems.
What to watch
Paper Resources
📖 Reader Mode
~2 min readAbstract:Data centers and their compute nodes require accurate and flexible digital twins capable of modeling the complex interplay of workloads, environmental parameters, and physical metrics. Current machine learning approaches for HPC and its telemetry typically rely on a static subset of anonymous, fixed-position sensor variables tailored to single tasks. Consequently, these models become obsolete when target tasks change or sensor metrics vary. We propose SeT-Diff, the first foundational model for compute node telemetry and time-series. Unlike rigid architectures, our diffusion-based approach conditions the generative process on each sensor's semantic description, decoupling the system dynamics from the structure of the dataset. Experiments on a real-world supercomputer dataset demonstrate a Mean Absolute Error (MAE) of 0.0470 on reconstruction tasks. SeT-Diff exhibits zero-shot permutation stability, maintaining accuracy with negligible degradation even when sensors are shuffled. A single pre-trained model effectively performs data imputation, forecasting, and virtual sensing - achieving a 0.033 MAE in thermal inference - making SeT-Diff an effective data-driven digital twin for HPC systems.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Performance (cs.PF) |
| Cite as: | arXiv:2607.22548 [cs.AI] |
| (or arXiv:2607.22548v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.22548 arXiv-issued DOI via DataCite |
Submission history
From: Giovanni B. Esposito [view email]
[v1]
Mon, 11 May 2026 00:43:50 UTC (105 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.