# Horizon-Aware Early Event Prediction for Tokamak Disruption Alarms

arXiv: https://arxiv.org/abs/2609.24443 · HTML: https://mlsanity.com/paper/2609.24443/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 28 references that are in the map.
- First posted on arXiv 21 Sep 2026.

## Builds on (most cited first, 7 of 7)

- [A Scalable Discrete-Time Survival Model for Neural Networks](https://mlsanity.com/paper/1805.00917/) (2018)
- [Deep Survival Machines: Fully Parametric Survival Regression and Representation Learning for Censored Data with Competing Risks](https://mlsanity.com/paper/2003.01176/) (2020)
- [Hybrid deep learning architecture for general disruption prediction across tokamaks](https://mlsanity.com/paper/2007.01401/) (2020)
- [An Introduction to Deep Survival Analysis Models for Predicting Time-to-Event Outcomes](https://mlsanity.com/paper/2410.01086/) (2024)
- [Temporal Label Smoothing for Early Event Prediction](https://mlsanity.com/paper/2208.13764/) (2022)
- [Dynamic Survival Analysis for Early Event Prediction](https://mlsanity.com/paper/2403.12818/) (2024)
- [Continuous Convolutional Neural Networks for Disruption Prediction in Nuclear Fusion Plasmas](https://mlsanity.com/paper/2312.01286/) (2023)

## Abstract (arXiv)

Reliable disruption prediction is essential for the safe operation of future tokamaks. Existing full-distribution survival methods model the complete residual time-to-disruption distribution, whereas operational decisions primarily depend on disruption risk within a finite prediction horizon. This mismatch motivates introducing Early Event Prediction (EEP) objectives into survival-based disruption prediction. We take Deep Survival Machines (DSM) as the full-distribution baseline and propose applying two established EEP methods to tokamak disruption prediction: Temporal Label Smoothing (TLS), which directly predicts disruption probability within a finite horizon, and survTLS, which additionally models the event-time distribution within that horizon. Using a common causal encoder, we compare these methods on DIII-D, Alcator C-Mod, and EAST. We distinguish threshold-free deadline ranking from validation-selected fixed-policy alarm performance and evaluate prediction horizons and encoder architectures. TLS achieves the best mean alarm performance on DIII-D and EAST, whereas all methods perform poorly on Alcator C-Mod. survTLS does not consistently outperform DSM, suggesting that directly learning horizon-level event probability is more effective than modeling detailed within-horizon event-time distributions in the present setting. Finally, the selected prediction horizons and encoder-ablation results vary across devices, reflecting differences in disruption characteristics.