Symbolic Regression on LLM-SRBench (official 239-problem split)
77Acc0.1 (%)PTB-Search
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PTB-SearchBackbone=DeepSeek-V4, Calls/prob.=1002026.07 | 77 | — | 9.24 | |
| PTB-SearchBackbone=Llama-3.1-8B, Calls/prob.=1002026.07 | 73.2 | — | 1.3 | |
| LLM-SRBackbone=Llama-3.1-8B, Calls/prob.=1,0002026.07 | 49.2 | 19.7 | — | |
| LaSRBackbone=Llama-3.1-8B, Calls/prob.=1,0002026.07 | 38.4 | 4.6 | — | |
| SGABackbone=Llama-3.1-8B, Calls/prob.=1,0002026.07 | 3.4 | 1.3 | — | |
| Direct promptingBackbone=Llama-3.1-8B, Calls/prob.=1,0002026.07 | 0.8 | 1.7 | — |