Researchers introduce SEER, a framework that improves time series forecasting by using prediction errors to refine event retrieval and causal reasoning. It outperforms existing models across six benchmarks.
Carried by: arXiv, HF Daily Papers. First seen: .