Research map / Large language models / Agents, code and math
SR-Scientist: Scientific Equation Discovery With Agentic AI
Shijie Xia, Yuhan Sun, Pengfei Liu
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 |
|---|---|---|---|
| Replace LLM-SR pipeline with SR-Scientist agent (GPT-OSS-120B) | GPT-OSS-120B · LSR-Synth (overall) | Acc_0.01: 28.16 -> 63.57 | Improved |
| Replace LLM-SR pipeline with SR-Scientist agent (Qwen3-Coder-480B) | Qwen3-Coder-480B-A35B · LSR-Synth (overall) | Acc_0.01: 41.08 -> 49.09 | Improved |
| Replace LLM-SR pipeline with SR-Scientist agent (GPT-OSS-20B) | GPT-OSS-20B · LSR-Synth (overall) | Acc_0.01: 33.33 -> 42.64 | Improved |
| Replace LLM-SR pipeline with SR-Scientist agent (Qwen3-Coder-30B, no RL) | Qwen3-Coder-30B-A3B · LSR-Synth (overall) | Acc_0.01: 24.55 -> 32.30 | Improved |
| Add RL training to SR-Scientist | Qwen3-Coder-30B-A3B · LSR-Synth (overall) | Acc_0.01: 32.30 -> 40.92 | Improved |
| Add RL training to SR-Scientist | Qwen3-32B · LSR-Synth (overall, Qwen3-32B) | Acc_0.01: 29.46 -> 39.53 | Improved |
| Add RL training to SR-Scientist | Qwen3-14B · LSR-Synth (overall, Qwen3-14B) | Acc_0.01: 18.60 -> 23.26 | Improved |
| Add RL training to SR-Scientist | Qwen3-8B · LSR-Synth (overall, Qwen3-8B) | Acc_0.01: 15.50 -> 20.93 | Improved |
| Use log-linear reward instead of stepwise reward in RL (ablation) | SR-Scientist + RL · LSR-Synth (overall) | Acc_0.01: 37.21 -> 40.92 | Improved |
| Add data analyzer tool (data analysis) (ablation) | SR-Scientist · LSR-Synth (overall) | Acc_0.01 | Improved |
| Add experience buffer of best equations (ablation) | SR-Scientist · LSR-Synth (overall) | Acc_0.01 | Improved |
| Increase maximum turns per iteration from 10 to 25 (ablation) | SR-Scientist · LSR-Synth (overall) | Acc_0.01 | Improved |
Closest papers by meaning
- LLM-SR: Scientific Equation Discovery via Programming with Large Language Models (ICLR 2025 · similarity 0.89)
- Agentic AI Scientists Are Not Built For Autonomous Scientific Discovery (NeurIPS 2026 · similarity 0.89)
- AI-Researcher: Autonomous Scientific Innovation (NeurIPS 2025 · similarity 0.88)
- LLM-SRBench: A New Benchmark for Scientific Equation Discovery with Large Language Models (ICML 2025 · similarity 0.88)
- ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows (ICLR 2026 · similarity 0.88)
- SciCode: A Research Coding Benchmark Curated by Scientists (NeurIPS 2024 · 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.
