Benchmarks
347 sources · 1,258 models × 3,038 benchmarks · 158,912 scoresSLDBench evaluates whether language-model agents can discover predictive scaling laws across eight scientific tasks. This suite-level row defines the benchmark family and has no direct score observations; Atlas records the official task-mean R² aggregate on its child benchmark.
No signal or difficulty estimate. Held out of the fit by design.
This is a benchmark-family page. Scores are recorded on the component benchmarks below rather than directly on the parent, because the components use different score scales and directions.
| Metric | Models | Leader | Direction | Date |
|---|---|---|---|---|
| SLDBench Task-Mean R² | 13 | GPT-50.7 | higher is better | 2026-01-14 |