Search papers, labs, and topics across Lattice.
This paper introduces AutoSR, a fully automated system for Symbolic Regression that enhances traditional methods by integrating a comprehensive Research State, which retains the scientific context and reasoning behind each candidate equation. By employing progressive-widening Monte Carlo tree search (PW-MCTS), AutoSR effectively navigates competing investigations while preserving a detailed record of motivations and computational evidence. The system successfully recovers algebraically equivalent relations across various benchmark challenges, including previously unsolved problems, demonstrating its potential to advance automated scientific inquiry beyond mere equation fitting.
AutoSR not only recovers complex equations but also retains the entire scientific rationale behind them, revolutionizing how we approach symbolic regression.
We introduce Automatic Symbolic Regression (AutoSR), a fully automated system that instantiates Research-Space Symbolic Regression by searching persistent scientific investigations rather than isolated equations. Finite, noisy data often yield numerically competitive expressions that imply very different behavior outside the observed regime, making numerical fit and syntactic complexity insufficient measures of scientific credibility. Existing approaches largely focus on improving expressions, yet the search typically retains little beyond the resulting formula and score, losing the scientific record, such as motivations and probes, that inform what to try next. AutoSR preserves this record in a \textbf{Research State}, coupling each candidate equation with the reasoning, computational evidence, and independent review developed along its branch. Proposer--reviewer agents develop these states under progressive-widening Monte Carlo tree search (PW-MCTS), which allocates computation across competing investigations, while the accumulated research record is ultimately synthesized into a final report that explains the leading relation and the basis for its selection. Across nine selected challenges from two benchmark suites, AutoSR recovers algebraically equivalent relations in every case, including three cp3-bench problems that no published system recovers and six structurally diverse LSR-Transform problems. Overall, AutoSR extends symbolic regression from equation-level search toward automated scientific investigation, allowing scientific knowledge and accumulated evidence to shape both what is explored and how the resulting equation is justified.
