Research map / Large language models / Agents, code and math
Deliberate Evolution: Agentic Reasoning for Sample-Efficient Symbolic Regression with LLMs
Xinyu Pang, Zhanke Zhou, Xuan Li, Fangrui Lv, Shanshan Wei, Sen Cui, Bo Han, Changshui Zhang
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 diagnostic tools (data profiler, residual, dimensional) (ablation) | Deliberate Evolution (Qwen3-4B) · LSR-Synth Physics | NMSE: 2.52e-2 -> 4.37e-4 | Improved |
| add reflective memory (ablation) | Deliberate Evolution (Qwen3-4B) · LSR-Synth Physics | NMSE: 1.34e-3 -> 4.37e-4 | Improved |
| use adaptive operator selection instead of fixed Refine (ablation) | Deliberate Evolution (Qwen3-4B) · LSR-Synth Physics | NMSE: 8.69e-3 -> 4.37e-4 | Improved |
| use adaptive operator weights instead of uniform sampling (ablation) | Deliberate Evolution (Qwen3-4B) · LSR-Synth Physics | NMSE: 1.02e-2 -> 4.37e-4 | Improved |
| add stagnation monitor (ablation) | Deliberate Evolution (Qwen3-4B) · LSR-Synth Physics | NMSE: 7.69e-4 -> 4.37e-4 | Improved |
| add diagnostic tools (ablation) | Deliberate Evolution (Qwen3-4B) · LSR-Synth Physics | Acc0.01 (%): 4.55 -> 15.91 | Improved |
| use Deliberate Evolution with 40% sample budget vs LASR | Deliberate Evolution (Qwen3-4B) · LSR-Transform | NMSE: 1.83e-1 -> 1.15e-1 | Improved |
| use Deliberate Evolution vs LASR | Deliberate Evolution (Qwen3-4B) · LSR-Transform | Acc0.01 (%): 30.91 -> 50.45 | Improved |
| use Deliberate Evolution under 1% Gaussian noise | Deliberate Evolution (Qwen3-4B) · LSR-Transform noise | NMSE: 2.46e-1 -> 1.79e-1 | Improved |
| use Deliberate Evolution on real-world stress-strain | Deliberate Evolution (Qwen3-4B) · Stress-Strain OOD | NMSE: 6.34e-1 -> 2.98e-1 | Improved |
| use Deliberate Evolution on Llama3.1-8B Chemistry accuracy | Deliberate Evolution (Llama3.1-8B) · LSR-Synth Chemistry | Acc0.01 (%): 16.67 (LLM-SR) -> 11.11 | Hurt |
Closest papers by meaning
- FunctionEvolve: Structure-Guided Symbolic Regression with LLMs (NeurIPS 2026 · similarity 0.90)
- Agentic-imodels: Evolving agentic interpretability tools via autoresearch (NeurIPS 2026 · similarity 0.88)
- Symbolic Regression with a Learned Concept Library (NeurIPS 2024 · similarity 0.88)
- SE-Agent: Self-Evolution Trajectory Optimization in Multi-Step Reasoning with LLM-Based Agents (NeurIPS 2025 · similarity 0.88)
- MetaSymNet: A Tree-like Symbol Network with Adaptive Architecture and Activation Functions (AAAI 2025 · similarity 0.87)
- LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling (NeurIPS 2026 · similarity 0.87)
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.
