Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
2026-08-10 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors introduce the Model Discovery Agent (MDA), which helps figure out how things work by combining a language model with techniques from statistics and experiment design. MDA suggests possible explanations for how systems operate, then plans smart experiments to test and improve these ideas, even when the true system might be initially unknown. They test MDA on tasks from physics, chemistry, and biology, showing it can learn accurate models using fewer experiments and better predict the results of actions not yet tried. Their approach iteratively refines the understanding by focusing on the parts that don’t yet make sense.
causal modelinterventional predictionlarge language modelBayesian inferencesequential Monte Carlosimulation-based inferencevalue-of-informationexperiment designmodel discoverymechanistic modeling
Authors
Kevin Murphy
Abstract
Predicting the answer to interventional ``what if'' questions --- the outcome of an action never taken --- requires a \emph{mechanistic}, causal model, not a curve fit; and learning such a model requires \emph{experiments}, because passive data leaves its mechanisms unidentified. Experiments are expensive, so the central problem is \emph{data efficiency}. We present the Model Discovery Agent (MDA), which couples a large language model (LLM), used as a \emph{proposer} of candidate structures, with standard Bayesian machinery --- sequential Monte Carlo (SMC) for parameter and structure posteriors, simulation-based inference (SBI) for intractable likelihoods, and value-of-information (VoI) for experiment design --- to discover latent mechanistic world models from few interventions. MDA operates in the M-open setting: when the truth lies outside the current hypothesis class, a predictive check flags the inadequacy and the proposer expands the hypothesis space with a new model whose parameters are then identified by designed experiments. We show that \emph{discovery and design reinforce}: the design step identifies the mechanism the discovery step proposes, and the identified mechanism improves predictions, enabling further discoveries from the remaining unexplained residuals. On three different benchmarks --- covering physics (\DPbench, \citep{wiemann2026discoverphysics}), chemistry (\CHEMbench, \citep{kabra2026autoscilab}) and biology (\HHbench, a new partially observed single-neuron electrophysiology benchmark we create) --- we show that MDA sets a new SOTA in terms of data-efficient model learning and reliable interventional forecasting ability.