Reproduce “What If TSF: Reframing Time Series Forecasting as Scenario-Guided Multimodal Forecasting” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jinkwan Jang, Hyunbin Jin, Hyungjin Park, Kyubyung Chae, Taesup Kim
WIT introduces a benchmark testing whether time-series forecasts change appropriately when future scenarios change. It holds history and domain background fixed, comparing scenario-conditioned forecasts with neutral forecasts in factual, relational, and matched-case evaluations. Experiments report a 14.92 ± 0.64% reduction in GPT-4o’s aggregate forecast error versus its neutral forecasts. But this does not establish sensible responses to alternatives.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.