Pub-AI: AI in Science

Latent Dynamics · 2026-09-30

NeuronDiscover: Agent-in-Twin for Mechanistic Discovery in Neuronal Microenvironments with World Action Models

Haowei Xu, Wanyi Fu, Hongbin Han, Zhaoheng Xie

arXiv:2609.35338PDF

Auto-summarized: this summary was generated by a language model from the paper’s text and has not been reviewed by an editor. Check the paper before relying on it. Benchmark numbers appear only after a human has verified them.

TL;DR

NeuronDiscover is an Agent-in-Twin framework for mechanistic discovery in neuronal microenvironments under “twin confounding” (mechanism change vs computational twin error). A shared mechanism-grounded World Action Model (WAM) drives prediction, intervention proposals, and observation design; independently adjudicated outcomes revise a scoped Mechanism–Intervention–Observation–Outcome (MIOY) graph, compiling supported relations into executable programs with discrepancy-adjusted acceptance bounds. Evaluated in simulated transport worlds and donor-disjoint current-clamp recordings.

Why it matters

The authors argue predictive accuracy alone cannot distinguish real mechanistic changes from errors in a computational twin under sparse observations. NeuronDiscover targets this by explicitly modeling mechanism–discrepancy beliefs and designing experiments whose outcomes independently adjudicate and revise a scoped discovery graph. This aims to produce certified mechanistic relations (not just good predictions) for neuronal microenvironment processes from both simulated worlds and transfer to recordings.

Method

  • Formalizes twin confounding and uses a joint mechanism–discrepancy belief to guide experiment selection so observations separate physical explanations from plausible twin error.
  • Uses a mechanism-grounded World Action Model (WAM) that answers forward, inverse, and sensing queries with uncertainty; predicted programs include observation design and validity conditions.
  • After frozen program selection, independently adjudicated outcomes revise a scoped MIOY graph; supported relations compile into Executable Mechanistic Programs with discrepancy-adjusted acceptance bounds.

Limitation

Lemma 2 supplies only a conditional implemented-level guarantee, and it is conditional on assumed per-route error bounds; the paper notes conditional-route disagreement is an empirical diagnostic and neither certifies nor proves small per-route error bounds for the full model.

Abstract (from arXiv)

Mechanistic discovery in neuronal microenvironments requires interventions and measurements that separate competing explanations of solute transport and neuronal response. Predictive accuracy cannot settle the question: a real mechanistic change and an error in the computational twin leave the same signature in sparse observations. We formalize this twin confounding and reason over a joint mechanism--discrepancy belief, designing experiments that separate the two. NeuronDiscover is an Agent-in-Twin framework whose shared, mechanism-grounded World Action Model (WAM) couples prediction, intervention proposals, and observation design; independently adjudicated outcomes revise a scoped Mechanism--Intervention--Observation--Outcome (MIOY) graph, whose supported relations compile into executable programs carrying discrepancy-adjusted acceptance bounds. We evaluate on simulated brain-fluid tracer-transport worlds adjudicated by an independently frozen finer-mesh reference solver, and on donor-disjoint public current-clamp recordings of cortical neurons. Counting only relations that reach a certified terminal status, and scoring abstentions as unresolved for every method, at a matched budget of 16 experiments over 32 source units NeuronDiscover resolves 4.0 relations per assigned world against 3.4 for the strongest baseline and 3.2 without graph revision, at 5% false support and 82% scope accuracy. Joint mechanism--discrepancy acquisition resolves 3.8 relations versus 2.9 for plug-in expected information gain; discrepancy-adjusted verification lowers accepted-program failure from 15% to 9% at 60% acceptance coverage; and transfer to the recordings yields 1.94 versus 1.53 relations per assigned world. Correctness is adjudicated within declared model worlds and archival recordings.

Related papers