Research map / Large language models / Agents, code and math
Influence-Guided Symbolic Regression: Scientific Discovery via LLM-Driven Equation Search with Granular Feedback
Evgeny Saveliev, Samuel Holt, Nabeel Seedat, David L. Bentley, Jim Weatherall, Mihaela van der Schaar
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add influence-guided pruning feedback (vs basic val-MSE feedback) with MCTS (ablation) | IGSR · Lung Cancer (with Chemo. & Radio.) | NMSE: 0.293 -> 0.000787 | Improved |
| add MCTS search to iterative refinement with influence feedback (ablation) | IGSR · Lung Cancer (with Chemo. & Radio.) | NMSE: 0.626 -> 0.000787 | Improved |
| add both MCTS and influence feedback (ablation) | IGSR · Lung Cancer (with Chemo. & Radio.) | NMSE: 4.85 -> 0.000787 | Improved |
| replace deterministic pruning with LLM agent pruning (IGSR-Agent) | IGSR · LLM-SRBench ID NMSE average rank | average rank: 1.35 -> 2.41 | Hurt |
| replace deterministic pruning with LLM agent pruning (IGSR-Agent) | IGSR · LLM-SRBench Term Recall | average rank: 1.99 -> 1.98 | No clear effect |
| IGSR vs best baseline ICL, Lung Cancer | IGSR · Lung Cancer | MSE: 0.0557 -> 5.64e-5 | Improved |
| IGSR vs ICL on Warfarin PK | IGSR · Warfarin PK | MSE: 0.784 -> 0.565 | Improved |
| add influence scores to iterative ICL baseline | IGSR · Lung Cancer (with Chemo. & Radio.) | MSE: 63.3 -> 0.0141 | Improved |
| IGSR vs black-box RNN | IGSR · Lung Cancer (with Chemo. & Radio.) | MSE: 0.013 -> 0.0141 | No clear effect |
| increase RNA Polymerase feature count 10 to 263 | IGSR · RNA Polymerase runtime | runtime: approx 25% increase | No clear effect |
Closest papers by meaning
- FunctionEvolve: Structure-Guided Symbolic Regression with LLMs (NeurIPS 2026 · similarity 0.90)
- LLM-SR: Scientific Equation Discovery via Programming with Large Language Models (ICLR 2025 · similarity 0.89)
- MOOSE-Chem2: Exploring LLM Limits in Fine-Grained Scientific Hypothesis Discovery via Hierarchical Search (NeurIPS 2025 · similarity 0.88)
- Symbolic Regression with a Learned Concept Library (NeurIPS 2024 · similarity 0.87)
- Improving Monte Carlo Tree Search for Symbolic Regression (NeurIPS 2025 · similarity 0.86)
- Ab Initio Nonparametric Variable Selection for Scalable Symbolic Regression with Large $p$ (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
