Research map / Large language models / AI and society
Foundation Models for Scientific Discovery: From Paradigm Enhancement to Paradigm Transition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
A position paper arguing that foundation models are moving science toward a new paradigm, proposing a three-stage framework (meta-scientific integration, hybrid human-AI co-creation, autonomous discovery) and reviewing FM applications across experimental, theoretical, computational, and data-driven science.
Problem
Whether foundation models merely enhance existing scientific methodologies or are redefining how science is conducted, and what risks and research directions follow from FM-enabled discovery.
Approach
Frames FM impact on science as a progressive transition across three stages defined by autonomy and task scope, and maps existing FM applications onto the four classical scientific paradigms to argue for a emerging fifth paradigm.
Key results
- Proposes a three-stage framework (meta-scientific integration, hybrid human-AI co-creation, autonomous scientific discovery) for FM-driven scientific evolution.
- Reviews FM applications across experimental, theoretical, computational, and data-driven paradigms, and identifies four risk dimensions: bias, hallucination/misinformation, reproducibility, and accountability.
Limitations
- No empirical evaluation; claims rest on conceptual argument and literature review.
- Framework stages are described qualitatively with blurry boundaries between them.
Experimental setup
Builds on: foundation models AlphaFold FunSearch Logic-LM Coscientist
Closest papers by meaning
- AI-Researcher: Autonomous Scientific Innovation (NeurIPS 2025 · similarity 0.88)
- Towards Scientific Discovery with Generative AI: Progress, Opportunities, and Challenges (AAAI 2025 · similarity 0.87)
- ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows (ICLR 2026 · similarity 0.87)
- Specialized Foundation Models Struggle to Beat Supervised Baselines (ICLR 2025 · similarity 0.86)
- MOOSE-Chem2: Exploring LLM Limits in Fine-Grained Scientific Hypothesis Discovery via Hierarchical Search (NeurIPS 2025 · similarity 0.86)
- SciCode: A Research Coding Benchmark Curated by Scientists (NeurIPS 2024 · similarity 0.85)
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.