Seq-Flow: Efficient Probabilistic Forecasting with Self-Rollout Error Control
Seq-Flow updates probabilistic forecasts from the previous distribution, cutting CRPS 65% with few flow steps.
Seq-Flow is a conditional flow model whose ODE moves samples from the previous forecast distribution to the updated one, enabling accurate few-step updates. Self-rollout training uses a moving-average copy to generate forecasts that initialize later updates, limiting accumulated error. On particle-accelerator beam spill forecasting it reduces CRPS by 65% under a few-NFE budget and stays competitive on fluid-dynamics tasks. Trained on at most four self-rollout updates, it remains accurate over more than 400 consecutive updates.
- Flow starts from the prior forecast rather than Gaussian noise.
- Self-rollout reuses generated forecasts as the next flow's source.
- CRPS falls 65% on beam-spill forecasting under few function evaluations.
- Accuracy holds beyond 400 updates after training on only four.
Full article220 words · extracted from arxiv.org · click to collapse
Many scientific forecasting tasks require updating a distribution over future trajectories as new observations arrive. Conventional diffusion and flow models generate each forecast from Gaussian noise, often at the cost of many sampling steps. Warm-start methods reuse earlier predictions to reduce this cost, but their models are not trained to perform the forecast update itself, which can compromise quality under few-step sampling. In this work, we introduce Seq-Flow, a conditional flow model whose ODE transports samples from the previous forecast distribution to the updated one. Because successive forecasts often differ only modestly, this transport starts from an informative distribution and can produce accurate updates with few flow evaluations. Recursive reuse also creates a challenge: errors in one forecast become errors in the initial states of subsequent flows. We address this with self-rollout training, in which a moving average copy of the model generates forecasts that initialize later training updates. Unlike self-forcing methods, which reuse generated outputs as conditioning context, Seq-Flow reuses them as the source of the next flow. Experiments On particle-accelerator beam spill forecasting show Seq-Flow reduces CRPS by 65% under a few-NFE sampling budget, while remaining competitive with strong baselines on fluid-dynamics forecasting tasks. Although trained on self-rollouts of at most four updates, Seq-Flow remains accurate over more than 400 consecutive updates. Our code is available at https://github.com/Graph-COM/Seq-Flow.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2610.10440