Agentic-MatriBO: Capability-Compiled Agentic Bayesian Optimization for Materials and Chemical Design

Yu Zeng; Yuchuan Chen; Ziyue Zhang; Zijian Wei; Yanyi Su; Guojiang Zhao; Hongshuai Wang; Zhifeng Gao

cs.LGOct 13, 2026

Agentic-MatriBO: Capability-Compiled Agentic Bayesian Optimization for Materials and Chemical Design

Authors: Yu Zeng, Yuchuan Chen, Ziyue Zhang, Zijian Wei, Yanyi Su, Guojiang Zhao, Hongshuai Wang, Zhifeng Gao

Organizations: DP Technology, Beĳing, China

Abstract

Materials and chemical optimization must choose the next synthesis, characterization, or high-fidelity simulation from sequential observations under a limited budget. Large language models can use task knowledge to propose adjustments, but composing surrogates, acquisition functions, and search strategies is constrained by compatibility, parameter dependencies, and evidence requirements. We present Agentic-MatriBO, a capability-compiled agentic Bayesian optimization (BO) framework that represents BO components and selection conditions as composable capabilities. A language policy proposes BO recipes and candidate plans; a compiler checks proposals against the registered capability catalog; and a deterministic execution layer validates and commits candidates. Oracle observations are stored in research memory and inform later rounds. We compare Agentic-MatriBO with eight baselines on four offline finite candidate pools---COF, PCE10, Buchwald--Hartwig (BH), and Suzuki--Miyaura---using five seeds per condition; results are reported descriptively. Across the four tasks, Agentic-MatriBO has the best mean endpoint quality after orienting each objective so that higher is better. On COF, it reaches the tabulated optimum in all five seeds both in the end-to-end comparison and after shared initialization, where it also retains an advantage in cumulative quality. Its normalized area under the best-so-far quality curve during the BO stage (nAUC-BO) is substantially higher on COF, near parity with the best baseline on PCE10 and BH, and lower on Suzuki. COF ablations show that removing the capability compiler or evidence gate coincides with lower cumulative quality and fewer exact-optimum hits. The framework provides an inspectable interface for evidence-conditioned BO adaptation.

Figures & tables

Role	Outputs and permissions	Boundary
Policy reasoning layer	Interprets the task and evidence; proposes recipes, candidate plans, and probes	Does not directly commit candidate IDs or generate target observations
Deterministic BO execution layer	Compiles the catalog, validates proposals, generates a candidate shortlist, fuses scores, and commits a candidate	Validation covers only the actual execution path; it does not generate target values
Advisory reviewers	Return bounded task scores and public justification for given candidates	Cannot add candidates, commit independently, or veto candidates unilaterally
Experiment oracle	Returns target values and related observations for submitted IDs	Does not replace target values with language-generated values
Research memory	Stores, links, and retrieves results, hypotheses, and reflections	Does not authorize execution or treat explanations as observations

Table 1: Roles, permissions, and commitment boundaries.

Figure 1: Research loop and execution boundaries of Agentic-MatriBO. The three blocks separate evidence-guided policy adaptation, candidate decision, and retrospective memory; deterministic interfaces define their permissions. Capability compilation supplies the catalog, the policy proposes a recipe and candidate plan, execution validates and commits a legal candidate, and the oracle returns the observation. Observations and evidence judgments are stored for later retrieval, with proposal, candidate, and post-observation reviews applied at their respective stages.

Figure 2: From capability contracts to candidate commitment. (a) Registered capabilities use unified contracts; (b) the compiler combines task, state, and evidence to define admissible recipes; (c) the policy proposes a recipe and candidate plan, with PROBE returning evidence and STOP returning a report; and (d) the execution layer validates, scores, and commits a legal candidate before the oracle observation.

Task	Input space	Candidate pool	Objective / source
COF (Xe/Kr)	14 numerical descriptors	605 candidates	Maximize adsorption selectivity; LABO finite table with GCMC targets [ 21 , 24 ]
PCE10	4 composition variables	1,040 records (1,020 compositions)	Minimize photodegradation; LABO finite table
Buchwald–Hartwig	4 categorical variables	790 candidates	Maximize yield; Buchwald–sub4 table [ 25 ]
Suzuki–Miyaura	5 categorical variables	5,760 candidates	Maximize yield; reaction-condition table [ 26 ]

Table 2: Tasks, candidate pools, and observation sources.

Figure 3: End-to-end optimization performance on four tasks. (a) COF; (b) PCE10; (c) BH; (d) Suzuki. Curves and shaded regions show the five-seed mean incumbent and $\pm$ 1 s.d.; the horizontal axis includes 10 initial oracle evaluations plus 30 subsequent evaluations. Lower is better for PCE10, whereas higher is better for the other tasks. The gray region denotes initialization, the inset enlarges the BO stage, and the dashed line marks the tabulated optimum. Each method uses its own initialization procedure; the figure compares the complete running configurations.

Task	Metric	Best baseline	Baseline score	Agentic-MatriBO
COF	Best-so-far $\uparrow$	LMABO	17.326 $\pm$ 1.416	18.476 $\pm$ 0.000
	nAUC-BO $\uparrow$	LMABO	0.868 $\pm$ 0.076	0.970 $\pm$ 0.039
	EF40 $\uparrow$	LMABO	0.938 $\pm$ 0.077	1.000 $\pm$ 0.000
	Late nAUC $\uparrow$	LMABO	0.898 $\pm$ 0.075	0.991 $\pm$ 0.015
PCE10	Best-so-far $\downarrow$	CAKE	0.0034 $\pm$ 0.0040	0.00162 $\pm$ 0.00000
	nAUC-BO $\uparrow$	Reasoning BO	0.9963 $\pm$ 0.0062	0.9964 $\pm$ 0.0021

Table 3: Endpoint and cumulative metrics across four tasks. Values are mean

\pm

s.d. over five seeds. For each task–metric pair, the table reports the best baseline and Agentic-MatriBO. Boldface indicates better Agentic-MatriBO mean performance; a dagger indicates that the two displayed means round to the same value and is not a significance claim.

Figure 4: COF optimization with shared initialization. For a given seed, the five methods in this group share the same 10 initial candidates and observations, followed by 30 evaluations per method. Curves and shaded regions show the five-seed mean and

\pm

1 s.d.; the gray region denotes initialization and the inset enlarges the subsequent stage.

Method	Best-so-far $\uparrow$	nAUC-BO $\uparrow$	EF40 $\uparrow$	Late nAUC $\uparrow$
HEBO	16.550 $\pm$ 1.184	0.818 $\pm$ 0.032	0.896 $\pm$ 0.064	0.861 $\pm$ 0.010
CAKE	17.875 $\pm$ 0.531	0.823 $\pm$ 0.118	0.967 $\pm$ 0.029	0.868 $\pm$ 0.127
LMABO	13.618 $\pm$ 5.286	0.662 $\pm$ 0.224	0.737 $\pm$ 0.287	0.698 $\pm$ 0.256
BORA	15.953 $\pm$ 2.955	0.769 $\pm$ 0.107	0.863 $\pm$ 0.160	0.845 $\pm$ 0.148
Agentic-MatriBO	18.476 $\pm$ 0.000	0.923 $\pm$ 0.038	1.000 $\pm$ 0.000	0.982 $\pm$ 0.026

Table 4: COF comparison with shared initialization. Values are mean

\pm

s.d. over five seeds. The five methods in this group receive the same ten initial candidates; the complete summary is in Appendix A.2.

Figure 5: Grouped COF ablations. (a) Warm-start ablation: the complete system (agentic warm start) and random warm start; (b) component ablation: the complete system, No review, and fixed GP+UCB under shared random initialization; (c) Harness ablation: the complete system, Bare LLM, No Capability Compiler, No Evidence Gate, and LLM commitment under another shared initialization. Each condition uses five seeds with a total budget of 10+30; lines and shaded regions show the mean and

\pm

1 s.d., and the dashed line denotes the tabulated optimum. Each panel compares only against the complete system in its own group.

Ablation group	Condition	Best-so-far $\uparrow$	nAUC-BO $\uparrow$	Late nAUC $\uparrow$	$t_{95}$ $\downarrow$	Global optimum hits
Warm start	Full (agentic)	18.476 $\pm$ 0.000	0.970 $\pm$ 0.039	0.991 $\pm$ 0.015	13.4 [5/5]	—
	Random	18.476 $\pm$ 0.000	0.888 $\pm$ 0.048	0.967 $\pm$ 0.045	23.4 [5/5]	—
Component	Full	18.476 $\pm$ 0.000	0.888 $\pm$ 0.048	0.967 $\pm$ 0.045	23.4 [5/5]	5/5
	No review	18.300 $\pm$ 0.171	0.865 $\pm$ 0.082	0.963 $\pm$ 0.046	20.0 [5/5]	0/5
	Fixed GP+UCB	14.773 $\pm$ 3.573	0.696 $\pm$ 0.189	0.753 $\pm$ 0.192	35.0 [2/5]	—
Harness	Full	18.476 $\pm$ 0.000	0.923 $\pm$ 0.038	0.982 $\pm$ 0.026	19.6 [5/5]	5/5

Table 5: Grouped COF ablations. Values are mean

\pm

s.d. over five seeds; brackets in

t_{95}

give threshold-hit counts; — indicates that exact-optimum hits were not separately recorded for that control. Comparisons are within groups, and “Full” denotes the complete system. Full names and results are in Appendix A.3.

Figure 6: LLM sensitivity on COF. (a) Area over evaluations 10–40; (b) late-stage area over evaluations 20–40. Bars, points, and error bars show the five-seed mean, individual-seed values, and $\pm$ 1 s.d., respectively. The y-axis uses a truncated scale. This comparison includes the effect of the model on both initialization and subsequent strategy decisions.

Model	Init. best	Early nAUC (10–20)	nAUC-BO (10–40)	Late nAUC (20–40)	Optimum hits
DeepSeek 4.0 Flash	16.128 $\pm$ 2.777	0.9266	0.9698 $\pm$ 0.0395	0.9914 $\pm$ 0.0151	5/5
DeepSeek 4.0 Pro	17.281 $\pm$ 1.860	0.9560	0.9817 $\pm$ 0.0308	0.9946 $\pm$ 0.0083	5/5
Qwen3-Max	18.113 $\pm$ 0.000	0.9837	0.9910 $\pm$ 0.0063	0.9947 $\pm$ 0.0081	3/5
GPT-5.6-sol	18.113 $\pm$ 0.000	0.9849	0.9912 $\pm$ 0.0066	0.9944 $\pm$ 0.0070	5/5

Table 6: COF model comparison. Initialization best value, BO-stage area, and late-stage nAUC are reported as mean

\pm

s.d.; early nAUC is reported as a mean;

n=5

per condition. The complete summary is provided in Appendix A.3.

Figure 7: Original and anonymous IDs in zero-observation ranking. (a) COF; (b) PCE10; (c) BH; (d) Suzuki. Each row corresponds to a model. Blue points denote original IDs and orange points anonymous IDs; the horizontal axis is the fraction of the anchored ID entering the top 40 across 10 rankings, error bars are Wilson 95% intervals, and numbers are hit counts. The vertical dashed line is the random level of 26.7%. Models receive candidate IDs and descriptors directly, without the Agentic-MatriBO strategy or execution process.

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Dataset	Method	Best-so-far	nAUC-BO	EF40	Late-stage nAUC
COF $\uparrow$	Vanilla BO	$13.381\,\pm\,5.055$	$0.636\,\pm\,0.206$	$0.724\,\pm\,0.274$	$0.679\,\pm\,0.236$
	HEBO	$16.550\,\pm\,1.184$	$0.818\,\pm\,0.032$	$0.896\,\pm\,0.064$	$0.861\,\pm\,0.010$
	LABO	$15.564\,\pm\,3.148$	$0.683\,\pm\,0.203$	$0.842\,\pm\,0.171$	$0.739\,\pm\,0.168$
	CAKE	$17.004\,\pm\,3.289$	$0.847\,\pm\,0.164$	$0.920\,\pm\,0.178$	$0.877\,\pm\,0.166$
	LMABO	$17.326\,\pm\,1.416$	$0.868\,\pm\,0.076$	$0.938\,\pm\,0.077$	$0.898\,\pm\,0.075$
	BORA	$16.723\,\pm\,1.919$	$0.832\,\pm\,0.135$	$0.905\,\pm\,0.104$	$0.863\,\pm\,0.137$

Appendix

Table A1: Complete baseline comparisons across the four tasks. Values are mean

\pm

s.d. over five seeds per condition. Rows list one method per line; the final row of each task block reports the improvement of Agentic-MatriBO relative to the best baseline.

Dataset	Method	Best-so-far	nAUC-BO	EF40	Late-stage nAUC
COF $\uparrow$	Vanilla BO	$13.381\,\pm\,5.055$	$0.636\,\pm\,0.206$	$0.724\,\pm\,0.274$	$0.679\,\pm\,0.236$
	HEBO	$16.550\,\pm\,1.184$	$0.818\,\pm\,0.032$	$0.896\,\pm\,0.064$	$0.861\,\pm\,0.010$
	LABO	$17.209\,\pm\,2.718$	$0.803\,\pm\,0.163$	$0.931\,\pm\,0.147$	$0.860\,\pm\,0.148$
	CAKE	$17.875\,\pm\,0.531$	$0.823\,\pm\,0.118$	$0.967\,\pm\,0.029$	$0.868\,\pm\,0.127$
	LMABO	$13.618\,\pm\,5.286$	$0.662\,\pm\,0.224$	$0.737\,\pm\,0.287$	$0.698\,\pm\,0.256$
	BORA	$15.953\,\pm\,2.955$	$0.769\,\pm\,0.107$	$0.863\,\pm\,0.160$	$0.845\,\pm\,0.148$

Appendix

Table A2: Complete COF comparison with shared initialization. Values are mean

\pm

s.d. over five seeds per condition. Rows list one method per line; the final row of each task block reports the improvement of Agentic-MatriBO relative to the best baseline.

Ablation group	Condition	Best-so-far $\uparrow$	nAUC-BO $\uparrow$	EF40 $\uparrow$	Late nAUC $\uparrow$	$t_{95}$ $\downarrow$
Warm start	Full (agentic)	$18.476\,\pm\,0.000$	$0.970\,\pm\,0.039$	$1.000\,\pm\,0.000$	$0.991\,\pm\,0.015$	13.4 [5/5]
	Random	$18.476\,\pm\,0.000$	$0.888\,\pm\,0.048$	$1.000\,\pm\,0.000$	$0.967\,\pm\,0.045$	23.4 [5/5]
Component	Full	$18.476\,\pm\,0.000$	$0.888\,\pm\,0.048$	$1.000\,\pm\,0.000$	$0.967\,\pm\,0.045$	23.4 [5/5]
	No review	$18.300\,\pm\,0.171$	$0.865\,\pm\,0.082$	$0.990\,\pm\,0.009$	$0.963\,\pm\,0.046$	20.0 [5/5]
	Fixed GP+UCB	$14.773\,\pm\,3.573$	$0.696\,\pm\,0.189$	$0.799\,\pm\,0.194$	$0.753\,\pm\,0.192$	35.0 [2/5]
Harness	Full	$18.476\,\pm\,0.000$	$0.923\,\pm\,0.038$	$1.000\,\pm\,0.000$	$0.982\,\pm\,0.026$	19.6 [5/5]

Appendix

Table A3: Complete grouped COF ablation results. Each condition uses five seeds; brackets after

t_{95}

report n/5, the number of seeds reaching the threshold out of five.

Model	Best init. value	Early nAUC (10–20)	nAUC-BO (10–40)	Late-stage nAUC (20–40)	Exact-optimum hits
DeepSeek 4.0 Flash	16.128 $\pm$ 2.777	0.9266	0.9698 $\pm$ 0.0395	0.9914 $\pm$ 0.0151	5/5, latest at evaluation 28
DeepSeek 4.0 Pro	17.281 $\pm$ 1.860	0.956	0.9817 $\pm$ 0.0308	0.9946 $\pm$ 0.0083	5/5, latest at evaluation 26
Qwen3-Max	18.113 $\pm$ 0.000	0.9837	0.9910 $\pm$ 0.0063	0.9947 $\pm$ 0.0081	3/5, reached at evaluations 15–23
GPT-5.6-sol	18.113 $\pm$ 0.000	0.9849	0.9912 $\pm$ 0.0066	0.9944 $\pm$ 0.0070	5/5, latest at evaluation 39

Appendix

Table A4: Complete COF model comparison. Each condition uses five seeds; the final column reports exact-optimum hits and evaluation indices.

Metric	Definition	Direction	Interpretation
Best-so-far	$I_{B}$	Task-dependent	The best observed objective value at the end of the budget, retaining the original physical units.
nAUC-BO	$\frac{1}{B-10}\int_{10}^{B}Q_{n}\,dn$	$\uparrow$	The normalized incumbent area over evaluations 10–40, measuring cumulative search quality during the BO stage.
EF40	$Q_{40}$	$\uparrow$	Full-table min–max-normalized endpoint quality; 1 denotes the best table value. Exact candidate-ID hits are counted separately.
Late-stage nAUC	$\frac{1}{40-20}\int_{20}^{40}Q_{n}\,dn$	$\uparrow$	The normalized incumbent area over evaluations 20–40, describing cumulative quality in the second half of the run.

Appendix

Table B1: Definitions and interpretation of the reported metrics.

Source label and reference	Implementation source and registered version	Implementation and adaptation scope
Vanilla BO [3–5]	In-house implementation	Numerical GP baseline
HEBO [29]	huawei-noah/HEBO , ee6112d39d1a	Finite-table adapter with unqueried-candidate mapping; quasi-random fallback after repeated failure (see B.2.5)
LABO [23]	Xinzhe309/LABO , c97d5615ebe8	Low-fidelity prediction and candidate selection through the upstream adapter
CAKE [19]	richardcsuwandi/cake , ece5cbc04b3e	Kernel search and finite-table selection through the upstream adapter
LMABO [21]	giang-n-ngo/lmabo , b1671f5ad79f	Adapted acquisition selection; parse failures raise explicit errors instead of silent UCB fallback
BORA [18]	Ablatif6c/bora-the-explorer , 83b2ebf86f4f	Region guidance and finite-table candidate handling

Appendix

Table B2: Baseline implementation sources, registered versions, and adaptation scope.

Task	Rows and unique input combinations	Processing and protocol boundary
BH	790 rows; 790 unique combinations	Fixed-product Buchwald_sub4 background with four variable categorical fields; the benchmark excludes the 35 cross-product training priors
Suzuki	5,760 rows; 5,760 unique combinations	Five categorical fields; the 29/5,731 train/test split in the data description is not equivalent to the 10+30 query budget reported here

Appendix

Table B3: BH and Suzuki candidate tables and data processing.

Figure C1: Strategy events and effective formulations across four tasks. (a) COF; (b) PCE10; (c) BH; (d) Suzuki. Each row corresponds to one seed. From top to bottom, the four bands represent strategy actions, the effective surrogate model, the effective acquisition function, and the effective search strategy. Hatching denotes a strategy draft, solid fill denotes a validated strategy proposal that was adopted, and a dark border denotes execution of an algorithmic tool; the last three formulation bands record the formulation that actually became effective after each event. The horizontal axis is the temporal order of audit events and does not correspond to oracle evaluation count.

Object	Current mechanism
Method hypothesis METHOD	States the claim, prediction, and falsification condition; does not include a domain contrast
Domain main-effect and interaction hypotheses	Must declare domain_contrast, including comparison arms, controlled factors, and a threshold; both types share the evaluator
Contrast evaluation	Uses observed records only; exact-control matching by default (optional tolerance). Judgments derive from group counts, thresholds, and direction
Per-hypothesis reflection	SUPPORT/REFUTE/INCONCLUSIVE, with evidence and a public rationale; missing an explicit verdict that may be cited is downgraded to uncertain; DIFFERENT propositions are handled under their own threshold rule when sufficient groups exist
Hypothesis state	Stores supporting/opposing citations, evaluation counts, and state changes, which feed the subsequent Policy
Playbook	Citations are resolved to experimental events in the current run and deduplicated by round and candidate; it may be activated by default when at least two supporting items exist and no resolved experimental counterevidence exists; counterevidence in unresolved citations is outside this judgment

Appendix

Table C1: Hypothesis, reflection, and research-memory mechanisms.

Stage	Declaration protocol	Current implementation
Atomic admission	The complete combination satisfies the conditions for the current round	Primarily checks atomic changes in CHANGE and retains atoms that can continue execution
Supporting material	The required support list is non-empty	Verifies list non-emptiness only; citation validity and premise entailment unverified
Portfolio	Each acquisition capability and the final surrogate are compatible and satisfy admission	Filters BLOCKED IDs and verifies interfaces; partial binding for NEEDS_EVIDENCE; zero-quota keys still checked
CHANGE $\to$ KEEP	Recheck the candidate plan under the final formulation	Candidate plan validated pre-guard; deferred components bypass re-validation
Candidate and fusion	Bounded candidates, valid reviews, deterministic submission	Decoupled quotas/slots; falls back to numerical utility if valid review dimensions are absent
Research feedback	Public state is available for subsequent retrieval	Typed hypotheses, reflections, and retrieval rules are saved

Appendix

Table C2: Declared protocol and current implementation.

Figure D1: Two-dimensional projections and incumbent-update examples. (a) COF; (b) PCE10; (c) BH; (d) Suzuki. The red solid, blue dashed, and teal dotted lines correspond to the proposed method, Vanilla BO, and Random Search, respectively; crosses and stars mark the first and final incumbents in the displayed run. PCE10 is a minimization objective, and the background shows direction-normalized quality; the gray region indicates that the sum of the two plotted components exceeds 1. The BH and Suzuki coordinates are categorical labels, and adjacency does not represent continuous physical distance; connecting lines indicate update order only.

Materials Discovery

See all 29 papers

Agentic Bayesian Optimization through Surrogate-Augmented Autoresearch
Jul 31, 2026Paul Brunzema, Louis Tiao, Nhat Le +3Surrogate-Assisted Optimization Bayesian Optimization
Meta · RWTH Aachen University, Germany
A Closer Look at Agentic BBO: Benchmarking LLM Agents for Black-Box Optimization
Oct 8, 2026Ming Chen, Rong-Xi Tan, Ke Xue +8Black-Box Optimization AI Agent Benchmarks
State Key Laboratory of Novel Software Technology, Nanjing University · School of Artificial Intelligence, Nanjing University
TagBO: LLM-Driven Task-Aware Graph Bayesian Optimization for Scientific Discovery
Oct 10, 2026Xinzhe Yuan, Zimu Zeng, Zhuo Chen +6Bayesian Optimization Large Language Model-Guided Optimization
Harbin Institute of Technology · Shanghai Artificial Intelligence Laboratory · Fudan University +4

Jul 31, 2026J/K move · Enter open · S save

Agentic Bayesian Optimization through Surrogate-Augmented Autoresearch

Paul Brunzema, Louis Tiao, Nhat Le, Kevin De Angeli+2

Meta · RWTH Aachen University, Germany

Surrogate-Assisted Optimization Bayesian Optimization Large Language Model-Guided Optimization

Open paper AskPDF arXiv

Bayesian optimization (BO) has become the standard tool for sample-efficient optimization and owes its efficiency to uncertainty-aware search driven by generic statistical priors. Richer domain priors can improve BO in principle, but encoding them through tailored kernels or problem structure is difficult and rarely done in practice. LLMs can help sidestep this difficulty by making informal priors from natural language, code, and documentation directly available to the optimizer. However, existing LLM-based BO methods either insert the LLM into a fixed role (surrogate, acquisition proxy, or configuration interface) or hand it broad control, sacrificing the systematic exploration that makes BO reliable. We introduce agentic Bayesian optimization: a paradigm in which an LLM agent is the central decision maker in the BO loop while a Bayesian backend provides the uncertainty-aware optimization substrate. The agent configures the problem, queries the backend, selects and commits evaluations, and can revise the optimization strategy during the run by tightening bounds, switching acquisition functions, proposing targeted evaluations, or even reframing the problem following new instructions or observed evidence. We instantiate this idea in Sara, a surrogate-augmented autoresearch agent, and lenz, a modular BoTorch-based backend that the agent can inspect and modify through a structured interface. Across synthetic and real-world benchmarks, Sara preserves the reliability of state-of-the-art BO without prior knowledge, outperforms LLM-based baselines, and uses natural-language priors to improve beyond standard BO. We further demonstrate the practical value of agentic BO in dynamic settings, where Sara reconfigures the full optimization problem on the fly as requirements change, a capability not previously available in standard BO.

Abstract

Materials Discovery

Explore similar work

Agentic Bayesian Optimization through Surrogate-Augmented Autoresearch

A Closer Look at Agentic BBO: Benchmarking LLM Agents for Black-Box Optimization

TagBO: LLM-Driven Task-Aware Graph Bayesian Optimization for Scientific Discovery