Artificial intelligence has demonstrated remarkable capability in predicting scientific properties, yet scientific discovery remains an inherently physical, long-horizon pursuit governed by experimental cycles. Most current computational approaches are misaligned with this reality, framing discovery as isolated, task-specific predictions rather than continuous interaction with the physical world. Here, we argue for embodied science, a paradigm that reframes scientific discovery as a closed loop tightly coupling agentic reasoning with physical execution. We propose a unified Perception-Language-Action-Discovery (PLAD) framework, wherein embodied agents perceive experimental environments, reason over scientific knowledge, execute physical interventions, and internalize outcomes to drive subsequent exploration. By grounding computational reasoning in robust physical feedback, this approach bridges the gap between digital prediction and empirical validation, offering a roadmap for autonomous discovery systems in the life and chemical sciences.
Figures & tables
Figure 1: From prediction-centric AI for science to embodied science. Left: the dominant AI4S workflow is human-orchestrated. Experts curate data sources into datasets, train task-specific models, and use predictions to inform the next experimental step; execution remains human-managed, yielding a loosely coupled loop across stages. Right: embodied science reframes discovery as a closed-loop process interaction with the physical world. Agentic embodied AI provides the enabling system layer that operationalizes this paradigm by integrating Perception – Language – Action – Discovery (PLAD): it grounds cognition in instrument-derived evidence (Perception), conducts reasoning and planning (Language), carries out embodied experimentation (Action), and internalizes outcomes into scientific insight (Discovery). The system can sustain long-horizon autonomous scientific discovery , executing iterative cycles beyond task-bounded, single-shot assistance.
∙ Strong language-based reasoning, but no coupling to embodied action; ∙ Operate primarily on curated datasets rather than raw signals from instruments; ∙ Remain physically disembodied, with experimental execution carried out by human scientists.
∙ Reliable embodied execution, but limited language-level reasoning and hypothesis formation; ∙ Operate directly on raw instrument signals using heuristic decision-making (e.g., Bayesian optimization); ∙ Enable autonomous execution within narrow task scopes and fixed procedural boundaries.
Single-step, instrument-bound
e.g., automated liquid handlers; robotic pipetting and weighing systems
Multi-step, protocol-driven automation
e.g., Chemputer [ 40 ] ; FLUID [ 41 ]
Table 1: Landscape of current AI4S approaches toward autonomous scientific discovery. Existing efforts cluster into two partial realizations: reasoning-centric but physically disembodied systems, and execution-centric but cognitively shallow platforms. Notably, neither paradigm achieves a sustained coupling between language-level scientific reasoning and embodied experimental action.
Figure 2: The PLAD loop for long-horizon autonolous scientific discovery. Perception transforms raw instrument signals into structured evidence. Language integrates foundation models with specialized knowledge and tools to support hypothesis formation, reasoning, and experimental planning. Action compiles plans into verifiable laboratory operations. Discovery internalizes outcomes into transferable scientific knowledge that shapes subsequent exploration across cycles.
Figure 3: Instantiating PLAD in scientific discovery. (a) Enzyme design: Perception integrates structural and functional measurements; Language forms structure–function hypotheses under constraints; Action executes library construction and screening; Discovery internalizes design rules and counterexamples for future cycles. (b) Chemical reaction optimization: Perception tracks spectroscopic and kinetic trajectories; Language performs mechanism-aware reasoning; Action executes targeted campaigns; Discovery internalizes mechanistic explanations and pathway-level constraints that generalize beyond parameter tuning.
Figure 4: Challenges and design resolution for sustaining long-horizon PLAD loops . (a) Reasoning over scientific data : Heterogeneous, noisy instrument signals require mechanism-aware interpretation; science-adapted LLMs, modality-aware encoders, and tool-grounded analysis support reliable reasoning. (b) Embodied execution reliability : Robust execution across diverse robotic embodiments requires large-scale simulation, trajectory learning, and sim-to-real transfer. (c) Long-horizon knowledge accumulation : Knowledge graphs convert fragmented experimental records into cumulative scientific insight across cycles. (d) Protocolization of scientific infrastructure : Agent-interpretable action–observation protocols standardize actions, observations, and failure states, enabling composable closed-loop workflows. (e) Safety and risk governance : Trustworthy long-horizon autonomy requires integrated safety gating, model-based risk assessment, and knowledge-based safety boundaries.
Large Language Models (LLMs) have shown significant potential in scientific discovery but struggle to bridge the gap between theoretical reasoning and verifiable physical simulation. Existing solutions operate in a passive "execute-then-response" loop and thus lacks runtime perception, obscuring agents to transient anomalies (e.g., numerical instability or diverging oscillations). To address this limitation, we propose EmbodiedAct, a framework that transforms established scientific software into active embodied agents by grounding LLMs in embodied actions with a tight perception-execution loop. We instantiate EmbodiedAct within MATLAB and evaluate it on complex engineering design and scientific modeling tasks. Extensive experiments show that EmbodiedAct significantly outperforms existing baselines, achieving SOTA performance by ensuring satisfactory reliability and stability in long-horizon simulations and enhanced accuracy in scientific modeling.
To unleash the full potential of AI for Science, we must untether the agents from a purely digital environment. The agent's ability to control and explore in real-world labs is essential because the physical lab remains foundational to scientific discovery. While some tasks can be performed on a computer (e.g., data analysis, running simulated experiments), Eureka moments could occur at any time while operating lab instruments (e.g., when a scientist notices unexpected clues, intuition may prompt a real-time course change). Although autonomous labs are on the rise, which expose programmable APIs to control scientific instruments via software, bridging the gap between increasingly powerful AI agents and automated lab equipment requires innovation that draws insights from computer systems. We propose a new paradigm called ``Experiment-as-Code (EaC) Labs,'' where a core concept is to encode experiments as declarative configurations that can be compiled down to device-level APIs. AI agents come up with hypotheses and experiments, written as an ensemble of declarative configurations. The systems layer performs program analysis, safety checks, resource assignment, and job orchestration. Finally, programmatic experimentation occurs via actuating the device APIs. This is a general stack that is science-, lab-, and instrument-independent, representing a novel synthesis across the physical, systems, and intelligence layers to unleash the next breakthrough in AI for Science.
Zhenning Yang, Yuhan Chen, Patrick Tser Jern Kon +5
A growing body of work pursues AI scientists capable of end-to-end autonomous scientific discovery. This position paper argues that although they already function as co-scientists, agentic AI scientists are not built for autonomous scientific discovery. We identify the following challenges in building and deploying autonomous AI scientists: (1) Problem selection is influenced by the McNamara fallacy; (2) Agents are built on large language models (LLMs) whose training corpora omit tacit procedural and failure knowledge of laboratory practice; (3) Preference optimisation during post-training compresses output diversity toward consensus; and (4) Most scientific benchmarks measure single-turn prediction accuracy and lack feedback from physical experiments back to the computational model. These challenges are not just questions of scale and scaffolding; they require revisiting fundamental design choices. To build truly autonomous AI scientists, we recommend the use of scientific simulations as verifiers for training, the design of persistent world models that represent the shifting objectives governing real investigations, the establishment of a centralized preregistration repository for all AI-generated hypotheses, and application driven by scientific need rather than tool affordance.
Harshit Bisht, Vinay Kumar, Kevin Maik Jablonka +2
Yardi School of Artificial Intelligence, Indian Institute of Technology Delhi · Laboratory of Organic and Macromolecular Chemistry (IOMC), Friedrich Schiller University Jena · Center for Energy and Environmental Chemistry Jena, Friedrich Schiller University Jena +3