fev-bench
A time series forecasting evaluation benchmark used to measure a model's predictive accuracy on marginal quantiles.
Timeline (last 90 days)
The 209M-parameter Timestral leads in the zero-shot model category on three benchmarks: GIFT-Eval, TIME and fev-bench.
fev-bench comes from the fev (Forecast Evaluation and Verification) ecosystem and focuses on providing reproducible evaluation pipelines and standardized data splitting methods.
Timestral-209M records 50.39 on the fev-bench SQL skill score, versus 49.42 for t0-gamma and 47.28 for Chronos-2.
PathFM achieves competitive marginal accuracy on the existing fev-bench and TIME benchmarks.
The research team proposed Forecast Path Score Bench (fps-bench), which adds energy scores, variogram scores, and downstream metrics on top of the fev-bench tasks.
All Facts (5)
The 209M-parameter Timestral leads in the zero-shot model category on three benchmarks: GIFT-Eval, TIME and fev-bench.
50%Unverifiedfev-bench comes from the fev (Forecast Evaluation and Verification) ecosystem and focuses on providing reproducible evaluation pipelines and standardized data splitting methods.
50%UnverifiedTimestral-209M records 50.39 on the fev-bench SQL skill score, versus 49.42 for t0-gamma and 47.28 for Chronos-2.
50%UnverifiedPathFM achieves competitive marginal accuracy on the existing fev-bench and TIME benchmarks.
50%UnverifiedThe research team proposed Forecast Path Score Bench (fps-bench), which adds energy scores, variogram scores, and downstream metrics on top of the fev-bench tasks.
50%