Metric-Aware Hybrid Forecasting for the CTF4Science Lorenz Challenge

2026-06-02

AI Summary

This work addresses the challenge of multi-objective evaluation in the CTF4Science Lorenz challenge—encompassing short-term prediction, long-term distribution matching, and full-trajectory reconstruction—by proposing a metric-aware hybrid forecasting architecture. The approach employs task-specific models tailored to each objective, revealing that a single unified model struggles to simultaneously satisfy all criteria. The methodology integrates synthetic data pretraining for denoising, fitting of the Lorenz ordinary differential equations, and trajectory shooting, augmented by a novel histogram tail replacement technique based on a synthetic Lorenz ensemble. The resulting system achieves a score of 83.83551 on the public leaderboard, which further improves to 83.85529 with minor refinements, demonstrating that precise alignment between tasks and dedicated models is crucial for maximizing overall performance.

Abstract

We describe our approach to the CTF4Science Lorenz challenge, a benchmark that mixes short-horizon forecasting, long-time distribution matching, and trajectory reconstruction across nine task pairs. The key discovery is that no single model family dominated all metrics. Instead, we built a metric-aware hybrid system that assigned a different predictor to each metric family: (1) synthetic-pretrained denoisers for full-trajectory reconstruction, (2) Lorenz ODE fitting and trajectory shooting for the first 20 forecast steps, and (3) histogram-tail substitution using synthetic Lorenz libraries for long-time evaluation. A representative mature submission from this system family scored 83.83551 on the public leaderboard, and a small follow-up stack of the same ideas reached 83.85529. We focus on the cleaner intermediate system because it captures the full method while remaining simple enough to reproduce and analyze, while the final submission can be understood as a conservative extension of the same backbone.

Problem

• Lorenz challenge
• short-horizon forecasting
• long-time distribution matching
• trajectory reconstruction
• metric-aware forecasting

Innovation

• metric-aware hybrid forecasting
• trajectory reconstruction
• Lorenz ODE fitting
• synthetic-pretrained denoiser
• histogram-tail substitution