Weekly sweeps
Tested every week against PySR, Operon and gplearn on public benchmark datasets: ten datasets, six from PMLB and four from the UCI repository. Every method gets the same data, the same time and the same seeds. This page shows the latest sweep. The weekly sweep is our own comparison; it is not the SRBench protocol.
These are the weekly sweeps of the earlier, own method: ten seeds, 1,800 s per fit, an own 75/25 split, and CPU cores that were not the same for every method. The current comparison runs monthly under the SRBench 2025 protocol.
The dataset feeding this page is rebuilt automatically on each release and served from the Ufinq cluster. When new results are available, they render below; otherwise the previous run is kept visible.
Loading
Fetching the latest run …
Read more
The harness that produces these results is described under Research § 2.6. The algorithms being benchmarked are described under Technology. To evaluate Ufinq on your own workload, run an automated analysis on Diafunc; to read models it has discovered on real data, browse the public analyses.
