---
title: 'CARTOGRAPH-A: A-Optimal Experiment Selector'
url: https://www.emergentmind.com/topics/cartograph-a
type: topic
---

# CARTOGRAPH-A: A-Optimal Experiment Selector

CARTOGRAPH-A is an experiment-selection rule within the broader CARTOGRAPH framework for autonomous scientific discovery. In that framework, an AI scientist is organized as a three-decision loop—**select**, **resolve**, and **refuse**—operating over a shared mechanism basis \(\Phi=\{\phi_1,\dots,\phi_p\}\), a model library \(\mathcal M=\{M_1,\dots,M_L\}\), and the controversial subspace on which library members disagree. CARTOGRAPH-A is the framework’s exact unresolved **A-optimal** acquisition rule: under a local linear-Gaussian bridge, it scores candidate experiments by the reduction they induce in posterior covariance trace restricted to the unresolved subspace, thereby refining simpler disagreement or projection heuristics into a covariance-aware experimental design criterion [2606.07576].

## 1. Conceptual setting and scientific objective

CARTOGRAPH is designed for the case in which an autonomous agent must identify scientifically relevant mechanisms from experimental behavior rather than by reading symbolic model coefficients directly. The paper formulates the true law as
\[
T(x)=\sum_{j=1}^{p} a_j^\star \phi_j(x),
\]
with the library \(\mathcal M=\{M_1,\dots,M_L\}\) retaining different subsets of the shared mechanism basis. The decisive object is the part on which library members disagree, termed the **controversial subspace** \(C\), with parameters \(a_C^\star\) [2606.07576].

Within this formulation, the central task is not generic exploration, but targeted elimination of unresolved scientific ambiguity. CARTOGRAPH therefore decomposes autonomous discovery into three distinct decisions. **Select** chooses the next experiment that best reduces the remaining ambiguity. **Resolve** decides when ambiguity has been closed enough to stop experimentation. **Refuse** determines when the current model library is structurally inadequate and should not be trusted [2606.07576].

A key theoretical distinction is that symbolic access is treated as a **coverage problem**, whereas behavioral access is treated as a **rank problem**. This suggests that experimental design should be driven by the geometry of unresolved disagreement, not merely by coverage of possible mechanisms. CARTOGRAPH-A is the paper’s most exact formulation of that design principle [2606.07576].

## 2. Unresolved subspace and behavioral inverse problem

In the behavioral setting, experiments produce disagreement matrices rather than direct access to mechanism coefficients. The paper writes the inverse problem as
\[
y = H a_C^\star + \varepsilon,
\]
where \(H\in \mathbb R^{n\times p_C}\) is the accumulated disagreement matrix. The current ambiguity is encoded by the **unresolved subspace**
\[
U_\tau = \operatorname{span}\{v_j:\sigma_j\le \tau\},
\]
where \(v_j\) are right singular vectors of the current disagreement matrix \(H_{\mathrm{cur}}\) with singular values \(\sigma_j\). In the exact case, \(\tau=0\) and
\[
U_0 = \ker(H_{\mathrm{cur}}).
\]
This makes \(U_\tau\) the subspace of directions that the present experimental record still cannot determine [2606.07576].

For a candidate experiment \(e\), CARTOGRAPH locally linearizes the predicted observables of each library member:
\[
g_{\ell,e}(z)\approx g_{\ell,e}(0)+J_{\ell,e}z.
\]
From these Jacobians it constructs pairwise disagreement blocks
\[
D_{ij,e} := J_{i,e}-J_{j,e},
\]
and stacks them into the experiment’s disagreement design
\[
H_e= \begin{bmatrix} D_{12,e}\\ D_{13,e}\\ \vdots\\ D_{(L-1)L,e} \end{bmatrix}.
\]
The scientific meaning of \(H_e\) is local but precise: it measures how strongly experiment \(e\) probes the currently unresolved directions of model disagreement [2606.07576].

The paper also gives a one-step gap-closure result,
\[
\dim(U)-\dim(U_{\mathrm{after}(e)})=\operatorname{rank}(H_e|_U),
\]
which implies that only the projected action of an experiment on the unresolved subspace matters for ambiguity reduction. This relation is the immediate precursor to both the raw unresolved projection score and CARTOGRAPH-A’s A-optimal refinement [2606.07576].

## 3. Mathematical definition of CARTOGRAPH-A

The simplest CARTOGRAPH selector is the isotropic unresolved-projection norm
\[
\operatorname{score}(e)=\|H_e U_\tau\|_F^2.
\]
The paper emphasizes that this is the **isotropic special case**. It measures how much a candidate experiment “hits” the unresolved subspace, but it does not account for anisotropic covariance or non-isotropic observation noise [2606.07576].

CARTOGRAPH-A introduces that missing covariance structure. Under the local linear-Gaussian bridge, it defines
\[
G_e = U_\tau^\top H_e^\top \Sigma_e^{-1} H_e U_\tau,
\]
and scores a candidate experiment by
\[
\operatorname{score}(e) = \operatorname{tr}(\Lambda_{\mathrm{cur}}) - \operatorname{tr}\!\big((\Lambda_{\mathrm{cur}}^{-1}+G_e)^{-1}\big).
\]
This is the framework’s **CARTOGRAPH-A** rule [2606.07576].

The corresponding local model assumes unresolved coordinates \(z\in\mathbb R^k\) obey
\[
z\sim \mathcal N(0,\Lambda_{\mathrm{cur}}), \qquad
y_e = H_eU_\tau z + \varepsilon_e, \qquad
\varepsilon_e\sim\mathcal N(0,\Sigma_e).
\]
Posterior precision then satisfies
\[
\Lambda_e^{-1}=\Lambda_{\mathrm{cur}}^{-1}+G_e,
\qquad
\Lambda_e=(\Lambda_{\mathrm{cur}}^{-1}+G_e)^{-1}.
\]
The criterion is therefore exactly the reduction in posterior covariance trace on the unresolved coordinates [2606.07576].

This is why the paper calls CARTOGRAPH-A the **exact unresolved A-optimal rule**. Classical A-optimality minimizes posterior covariance trace; CARTOGRAPH-A applies that principle only on \(U_\tau\), the subspace that remains scientifically unresolved. A plausible implication is that the method separates “what is still uncertain” from “what matters for experimental discrimination,” rather than optimizing over the full parameter space indiscriminately.

## 4. Relation to raw projection, EIG, and Box–Hill

Under isotropic observation noise, \(\Sigma_e=\sigma_e^2 I\), the paper shows that
\[
\operatorname{tr}(G_e)=\sigma_e^{-2}\|H_eU_\tau\|_F^2.
\]
Thus raw unresolved projection becomes exactly the **Fisher-information trace** on the unresolved subspace. In this limit, raw CARTOGRAPH is not an ad hoc heuristic but the isotropic Fisher-trace special case of the same local design logic [2606.07576].

The paper also derives a closed-form local expression for expected information gain:
\[
\mathrm{EIG}(e)=\tfrac12 \log\det\!\big(I+\Lambda_{\mathrm{cur}}^{1/2}G_e\Lambda_{\mathrm{cur}}^{1/2}\big).
\]
For small information gain,
\[
\mathrm{EIG}(e)\approx \tfrac12 \operatorname{tr}(\Lambda_{\mathrm{cur}}G_e).
\]
Accordingly, EIG is described as **first-order aligned** with CARTOGRAPH-A, but not identical to it: EIG is a log-det criterion, whereas CARTOGRAPH-A is a trace-reduction criterion [2606.07576].

For Box–Hill, the paper gives
\[
\mathrm{BH}(e)=\tfrac12\big[\operatorname{tr}(V_1^{-1}V_2)+\operatorname{tr}(V_2^{-1}V_1)\big]
+\tfrac12\Delta^\top(V_1^{-1}+V_2^{-1})\Delta,
\]
with \(\Delta=m_1(e)-m_2(e)\). Under shared isotropic noise and equal predictive covariances \(V_1=V_2=\sigma_e^2 I\), this collapses to
\[
\sigma_e^{-2}\|\Delta\|_2^2
\]
plus a constant. The contrast is therefore structural: Box–Hill remains a pairwise model-discrimination criterion, while CARTOGRAPH-A is an unresolved-subspace A-optimal design over the library’s disagreement geometry as a whole [2606.07576].

The paper’s broader theoretical message is that EIG and Box–Hill appear as **local comparators rather than global equivalents**. This suggests that CARTOGRAPH-A occupies an intermediate position: more principled than raw disagreement magnitude, but more specialized than global Bayesian design objectives because it explicitly restricts attention to unresolved coordinates [2606.07576].

## 5. Algorithmic role in the select–resolve–refuse loop

CARTOGRAPH iterates through a fixed decision procedure. First, it computes \(U_\tau\) from the accumulated disagreement matrix. If \(\dim(U_\tau)=0\), it **resolves**: ambiguity is closed. Otherwise, it chooses the next experiment \(e^\star\) by maximizing either the raw score \(\|H_eU_\tau\|_F^2\) or the CARTOGRAPH-A trace-reduction objective. After the experiment is executed, the disagreement matrix is updated and the system evaluates refusal diagnostics [2606.07576].

The refusal step is based on two quantities:
\[
\rho = \frac{\min_{\ell}\|y_{\mathrm{obs}}-f_{m_\ell}(\hat\theta_\ell)\|_2}{\|\phi(y_{\mathrm{obs}})\|_2},
\qquad
\mu=\mathrm{BIC}(m_{(2)})-\mathrm{BIC}(m_{(1)}).
\]
Identification is permitted only if
\[
\rho\le \delta
\quad\text{and}\quad
\mu\ge \mu_{\min}.
\]
If \(\rho>\delta\), the system **refuses** and revokes any tentative identification [2606.07576].

A central point in the paper is that refusal is **not** the same as predictive uncertainty. It is driven by **library-relative residual misfit**. This allows the system to behave dynamically: it may identify early, then retract the claim when further data reveal structural mismatch. The paper treats this revocation behavior as a governance property rather than a failure mode [2606.07576].

Within this loop, CARTOGRAPH-A has a sharply delimited function. It is a selector, not a stopping rule and not a refusal diagnostic. Resolve depends on closure of unresolved ambiguity; refuse depends on residual-based evidence of model-library inadequacy. The separation of these three decisions is one of the framework’s defining features [2606.07576].

## 6. Empirical performance and boundary cases

The paper’s main selection benchmark is a **structured nonlinear cascade**. At \(d=8\), it reports the replicated result **Cartograph-A vs raw Cartograph: 129W / 0T / 15L**, with \(p < 10^{-21}\). At \(d=16\), it reports **120W / 0T / 24L**, with \(p<10^{-16}\). By contrast, at \(d=2\) all methods tie completely, and at \(d=4\) the behavior is transitional or mixed [2606.07576].

These results support the paper’s stated interpretation: in higher dimensions, raw projection leaves information unused, whereas CARTOGRAPH-A exploits anisotropy in unresolved covariance. In low dimensions, there may be little room to improve on simpler unresolved-projection or disagreement-style heuristics. The paper also states that CARTOGRAPH-A tracks closed-form EIG very closely in this benchmark while being much cheaper than Monte Carlo BOED [2606.07576].

The low-dimensional pharmacokinetics benchmark is presented as a boundary case. There, raw Cartograph is reported as **\(1W/6T/0L\)** versus disagreement, and CARTOGRAPH-A is said to be identical to raw Cartograph on all seven truths; the one-sided sign test is not significant. In filtered EPA pharmacokinetics, the results are again near-tie or modest-gain: raw Cartograph versus disagreement is **\(1W/7T/0L\)**, and local T-opt versus disagreement is **\(2W/6T/0L\)** [2606.07576].

The paper explicitly treats these outcomes as theoretically expected rather than anomalous. Its stated scaling law says that in random-candidate settings, the useful unresolved fraction scales like \(k/d\), explaining when disagreement heuristics are sufficient and when CARTOGRAPH-A should outperform them. This suggests that CARTOGRAPH-A is most distinctive in **high-dimensional and anisotropic unresolved settings**, not as a universal dominance result [2606.07576].

## 7. Refusal behavior, retrospective audit, and significance

Although CARTOGRAPH-A is a selection rule, the paper’s broader significance depends on its embedding within the refuse mechanism. In an out-of-library pharmacokinetic benchmark, the framework tentatively identifies three mechanisms outside the library—**time-varying clearance**, **saturable elimination**, and **enterohepatic recirculation**—and then revokes those identifications when residuals cross the threshold \(\delta=0.25\). A perturbed in-library control remains below threshold and stays identified [2606.07576].

The paper also reports a retrospective audit of **40 positive claims** from the published A-Lab autonomous materials system. Among **36 confirmed claims**, the refuse guard passes **32** and flags **4**. Among **4 later-inconclusive claims**, it flags **4** and passes **0**. The paper contrasts this with an Rwp-only baseline that flags none of those inconclusive cases [2606.07576].

These findings matter for the interpretation of CARTOGRAPH-A because they situate it inside a broader doctrine of autonomous scientific conduct. The framework’s operative message is not only that experiment selection should reduce unresolved covariance efficiently, but also that an AI scientist should know when to stop and when to withhold belief. CARTOGRAPH-A supplies the exact unresolved-subspace A-optimal rule for the **select** decision, while the residual and BIC guards govern **resolve** and **refuse** at the system level [2606.07576].

In that sense, CARTOGRAPH-A is best understood not as a stand-alone acquisition function, but as the mathematically exact selector within a tripartite policy for autonomous discovery. Its contribution is strongest where unresolved disagreement is high-dimensional and structured; its restraint comes from the surrounding refusal machinery, which prevents efficient experiment selection from being mistaken for reliable scientific identification [2606.07576].

Source: https://www.emergentmind.com/topics/cartograph-a