Symbolic Regression on NewtonBench Hard
51.4Symbolic AccuracyLLM-AutoSciLab
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LLM-AutoSciLabED (Experiment Design)=true2026.05 | 51.4 | 66.7 | 0.404 | |
| PySRED (Experiment Design)=false2026.05 | 5.56 | 59.72 | 0.398 | |
| Bayesian OptimizationED (Experiment Design)=true2026.05 | 5.56 | 50 | 0.73 | |
| Bayesian Experimental DesignED (Experiment Design)=true2026.05 | 2.78 | 45.31 | 1.01 | |
| Code-assisted LLMED (Experiment Design)=true2026.05 | 2.78 | 0 | 4.897 | |
| LLM-onlyED (Experiment Design)=true2026.05 | 0 | 0 | 6.757 |