Mambo: Multifaceted Research Acronym

• Mambo is a polysemous research term that spans multiple fields, notably astronomy with bolometer arrays and high-redshift galaxy surveys.
• In astronomy, MAMBO enables 1.2 mm continuum mapping for source discovery and tracking dusty, star-forming galaxies using instruments like the IRAM 30 m telescope.
• Beyond astronomy, Mambo is applied in Bayesian optimization, materials ontology, systems security, robotics, and medical imaging, each defined by its specific context.

In arXiv usage, Mambo or MAMBO is not a single concept but a polysemous research label. Its best-known astronomical meaning is the Max-Planck Millimeter Bolometer camera or bolometer array on the IRAM 30 m telescope, used for 1.2 mm observations of dusty galaxies, protostellar condensations, and debris disks (Aravena et al., 2010). In later literature, the same string is reused for a Bayesian optimization algorithm, a materials-science ontology, a RISC-V dynamic binary instrumentation tool, a galaxy-and-AGN mock-catalogue workflow, and several machine-learning systems in medical imaging and OOD detection (Wang et al., 2022, Piane et al., 2024, Wichelmann et al., 2023, López-López et al., 2024). The term is therefore field-dependent, and its meaning is determined by disciplinary context.

1. Principal astronomical meaning: the Max-Planck Millimeter Bolometer

In astronomy, MAMBO denotes the Max-Planck Millimeter Bolometer instrument family at the IRAM 30 m telescope. In the $\epsilon$ Eridani debris-disk study, the paper specifies MAMBO-2 as a 117-channel bolometer camera operating over 210–290 GHz, with an effective frequency centered on 250 GHz, corresponding to $\lambda = 1.2~\mathrm{mm}$, an array footprint of $240'' \times 240''$, and a beam FWHM of $10.7''$ (Lestrade et al., 2015). In the Aquila Rift star-formation study, MAMBO is described as the Max-Planck Millimeter Bolometer array used on the IRAM 30 m telescope for 1.2 mm dust continuum mapping, with native angular resolution of about $11''$ (Maury et al., 2011).

Form	Domain	Meaning
MAMBO	mm/sub-mm astronomy	Max-Planck Millimeter Bolometer
MamBO	Bayesian optimization	Model Aggregation Method in Bayesian Optimization
MAMBO	materials semantics	Materials and Molecules Basic Ontology
MAMBO-V	systems security	RISC-V port of MAMBO DBI
Parrot Mambo	robotics/UAV	micro-UAV platform
MAMBO	medical imaging	MAMmography ensemBle mOdel

Operationally, MAMBO is used in chopped, scanned single-dish continuum mapping, with reductions performed in MOPSIC in the $\epsilon$ Eridani study and standard MOPSIC/GILDAS processing in the Aquila work (Lestrade et al., 2015, Maury et al., 2011). Its scientific role is usually source discovery or wide-field mapping rather than the subarcsecond localization provided by interferometers. In the COSMOS SMG work, MAMBO supplied the source-discovery step in a 1.2 mm blank-field survey, after which Bolocam, SMA, or CARMA follow-up resolved counterpart ambiguities (Aravena et al., 2010, Smolcic et al., 2012). This combination of survey depth, cold-dust sensitivity, and coarse single-dish beam made MAMBO central to the pre-ALMA identification of dusty high-redshift populations.

2. MAMBO-selected galaxies and the high-redshift dusty universe

One major astronomical usage of MAMBO is as a selection instrument for luminous dusty star-forming galaxies. In the COSMOS field, Bertoldi et al.’s MAMBO catalog yielded 15 sources at $\mathrm{S/N}>4$, of which five lacked significant 20 cm radio counterparts at the $\lesssim 30~\mu\mathrm{Jy}$ level. Follow-up SMA imaging of two such radio-faint sources showed that MM1 is a robust luminous SMG with a favored redshift around $z\sim 4.5$–4.8, while MM14 is a less secure but still plausible high-redshift candidate with a submm-to-radio lower limit of $z>3.5$. On that basis, the paper argued that up to $\lambda = 1.2~\mathrm{mm}$0 of the $\lambda = 1.2~\mathrm{mm}$1 mJy MAMBO sample in COSMOS could lie at $\lambda = 1.2~\mathrm{mm}$2, corresponding to a surface density of $\lambda = 1.2~\mathrm{mm}$3–$\lambda = 1.2~\mathrm{mm}$4 and a volume density of $\lambda = 1.2~\mathrm{mm}$5–$\lambda = 1.2~\mathrm{mm}$6 (Aravena et al., 2010).

The same methodological point appears in CARMA follow-up of bright COSMOS SMGs initially found by MAMBO and AzTEC. There, the authors showed that single-dish MAMBO positions were insufficient for secure counterpart assignment: for Cosbo-3, the true mm counterpart was not the radio source nearest the MAMBO centroid, but a farther northwestern radio source identified only after $\lambda = 1.2~\mathrm{mm}$7–$\lambda = 1.2~\mathrm{mm}$8 CARMA imaging. The three CARMA-localized systems had photometric redshifts of $\lambda = 1.2~\mathrm{mm}$9, $240'' \times 240''$0, and $240'' \times 240''$1, radio-based star-formation rates of order $240'' \times 240''$2, and infrared luminosities of order $240'' \times 240''$3 (Smolcic et al., 2012). This established MAMBO as a powerful discovery survey but a limited identification tool in crowded fields.

MAMBO-selected galaxies were also used as environmental tracers. In the COSMOS environment study, the authors found three statistically significant compact overdensities centered on MAMBO SMGs, with photometric redshifts consistent with the associated sources and lying in the range $240'' \times 240''$4–2.5. Their conclusion was that about 30% of the radio-identified bright SMGs in that redshift interval formed in galaxy density peaks, reinforcing the interpretation of at least a subset of MAMBO galaxies as signposts of proto-cluster assembly (0911.4297).

A related but later usage is MAMBO-9, shorthand for MMJ100026.36+021527.9. In its 2019 characterization, MAMBO-9 was presented as an unlensed dusty star-forming galaxy at $240'' \times 240''$5, resolved into a pair of galaxies separated by 6 kpc, with star-formation rates of $240'' \times 240''$6 and $240'' \times 240''$7, total molecular hydrogen gas mass of $240'' \times 240''$8, dust mass of $240'' \times 240''$9, and predicted halo mass $10.7''$0 (Casey et al., 2019). JWST and ALMA later refined this picture: at $10.7''$1 pc [CII] resolution, both components show velocity gradients, the dynamical mass ratio is roughly 1:5, a continuous bridge of moderately dust-obscured material links the pair, the preferred $10.7''$2 is roughly unity, and the system sits in an overdensity with 39 spectroscopically confirmed galaxies within $10.7''$3 cMpc (Akins et al., 8 Aug 2025). Across these studies, “MAMBO” denotes both the historical survey provenance of the source and, by extension, an extreme class of early dusty galaxy assembly.

3. Nearby debris disks and Galactic star formation

MAMBO was also used extensively for nearby cold-dust systems. In the $10.7''$4 Eridani study, a high-S/N 1.2 mm MAMBO image provided an independent check on the debated clumpy morphology first seen with SCUBA at $10.7''$5. The paper found that three of the four emission clumps and the two deep hollows agreed between the MAMBO and SCUBA maps within astrometric uncertainty, while the southeast clump was discrepant and may have been an artifact in the earlier SCUBA data. The same work derived a radial belt peak at $10.7''$6 and an intrinsic width $10.7''$7, corresponding to $10.7''$8 and therefore a belt narrower than the Kuiper Belt (Lestrade et al., 2015). In this context, MAMBO functioned as an independent millimeter imager for testing azimuthal structure and constraining belt thickness.

In Galactic star-formation studies, MAMBO provided the millimeter anchor for embedded protostellar populations. The Aquila Rift analysis used MAMBO 1.2 mm mapping of ten adjacent subfields, yielding a global mosaic of approximately $10.7''$9 and a catalog of 77 compact continuum sources. Combined with Herschel and Spitzer data, these MAMBO detections supported evolutionary classification in Serpens South and W40, including a Class 0 to Class I ratio in Serpens South of $11''$0–$11''$1 and an inferred Class 0 lifetime of $11''$2–$11''$3 yr. The same paper estimated star-formation-rate surface densities of roughly $11''$4–$11''$5 and argued that both protoclusters were undergoing bursts of clustered star formation (Maury et al., 2011). Here MAMBO’s value lay in tracing cold dust in dense envelopes and starless condensations that are inaccessible to shorter-wavelength surveys alone.

4. MamBO in optimization and state-space video modeling

Outside astronomy, MamBO is the Model Aggregation Method in Bayesian Optimization, introduced for high-dimensional, large-scale black-box optimization. The method considers optimization of

$11''$6

with $11''$7, and addresses three issues simultaneously: high dimension, large $11''$8, and surrogate uncertainty induced by embedding. Its pipeline partitions data into $11''$9 subsets, applies a subspace embedding $\epsilon$0 to each subset, fits a stochastic GP surrogate $\epsilon$1 on each embedded subset, computes Bayesian model weights $\epsilon$2, and aggregates the submodels into

$\epsilon$3

The paper derives an asymptotic bound for the aggregated predictor and proves convergence of the BO procedure under stated assumptions. Empirically, MamBO is reported as superior or comparable to REMBO, HesBO, TuRBO, ALEBO, and SAASBO on synthetic benchmarks, and in a Viola–Jones cascade-classifier tuning task over 22 stage thresholds it achieved less than 1% classification error after 220 iterations, outperforming the OpenCV default setting of 2.61% while being faster than several high-dimensional BO baselines (Wang et al., 2022).

A later reuse of the label is MamBOA, a video-recognition framework built around Mamba-style selective state-space models. MamBOA is described as a backbone-agnostic temporal module that interleaves consecutive feature representations so that the selective recurrence acts as a motion synthesizer. On Diving48, it achieved 85.02% Top-1 accuracy with an image-pretrained backbone and 86.24% Top-1 with a video-pretrained backbone processing the entire video in a single forward pass, while adding only $\epsilon$4 GFLOPs per feature pair (Çelik, 13 Jun 2026). The naming overlap is nominal rather than genealogical: MamBOA is a separate state-space video architecture, not an extension of Bayesian optimization.

5. MAMBO as an ontology for multiscale materials research

In materials informatics and semantic technologies, MAMBO denotes an ontology initiative for molecular and multiscale materials. The 2021 paper introduced MAMBO as the Materials And Molecules Basic Ontology (Piane et al., 2021). The 2022 follow-up described it as the Molecular and Materials Basic Ontology (Piane et al., 2022). The 2024 paper returned to Materials and Molecules Basic Ontology, emphasizing its role as a lightweight ontology for multiscale materials and applications (Piane et al., 2024). Across these variants, the target remains the same: formal organization of knowledge for materials whose relevant behavior depends on molecular structure, aggregation, supramolecular organization, and processing.

The core conceptual architecture is stable. Early papers revolve around classes such as Material, Structure, Material Property, Measurement, and Calculation, with relations such as has Structure, has Property, isMeasuredBy, and isCalculatedBy (Piane et al., 2021, Piane et al., 2022). The 2024 version reframes the same operational space around Material, Structure, Property, Simulation, and Experiment, maintaining the idea that structures and properties are linked to both computational and experimental workflows (Piane et al., 2024). The ontology is explicitly modular, intended to reuse concepts from EMMO, MDO, ChEBI, and CHMO, and implemented in OWL, with development in Protégé and public release on GitHub (Piane et al., 2021, Piane et al., 2024).

Methodologically, MAMBO is tied to problem-solving methods and competency-question design. Its intended use cases include retrieval of structured information on molecular materials, definition of complex workflows for modeling such systems, and integration of computational and experimental data for machine-learning and predictive-design pipelines (Piane et al., 2022, Piane et al., 2024). Structural representation is notably detailed: the ontology models StructuralEntity hierarchies including Atom, Particle, StructuralUnit, and MolecularSystem, and later introduces subclasses such as MolecularAggregate and Crystal (Piane et al., 2022, Piane et al., 2024). The liposome-in-water example used in the early papers illustrates the ontology’s focus on multicomponent nanoscale systems rather than only ideal crystals or isolated molecules (Piane et al., 2021, Piane et al., 2022).

6. Systems, security, robotics, medical imaging, and other contemporary acronyms

Several later arXiv papers reuse MAMBO in domain-specific acronyms unrelated to either the bolometer or the ontology. In systems security, MAMBO-V is a RISC-V port of the ARM-origin MAMBO dynamic binary instrumentation framework. It instruments arbitrary RV64GC binaries and acts as the trace-generation backend for Microwalk, enabling dynamic analysis of side-channel leakage in RISC-V cryptographic libraries. The paper highlights RISC-V-specific engineering issues such as short branch reach, gp/tp context handling, and software emulation of LR/SC atomic sequences (Wichelmann et al., 2023).

In survey cosmology, MAMBO is also expanded as Mocks with Abundance Matching in BOlogna, an empirical workflow for galaxy and AGN mock catalogues built on dark-matter lightcones. In its 2024 presentation, the proof-of-concept Millennium lightcone covers 3.14 deg$\epsilon$5 to $\epsilon$6 and is complete to at least $\epsilon$7. For the Euclid $\epsilon$8 Wide survey, the catalogue forecasts $\epsilon$9 galaxies, $\mathrm{S/N}>4$0 Type 1 AGN, and $\mathrm{S/N}>4$1 Type 2 AGN over 14679 deg$\mathrm{S/N}>4$2 (López-López et al., 2024). A 2026 FIR extension uses the same MAMBO mock to predict dust properties and infrared luminosity functions for Euclid-detectable galaxies, concluding that only the brightest Euclid sources will be directly detectable in present FIR data and that areas at least 30 times larger than the 3.14 deg$\mathrm{S/N}>4$3 mock are required for statistically robust $\mathrm{S/N}>4$4 environmental constraints (Collaboration et al., 13 Mar 2026).

In UAV perception, the Parrot Mambo is the micro-drone platform used for indoor Roomba tracking. The system collected 17,983 frames, used YOLOv4 for semi-automated labeling, and trained both YOLOv4 and Mask R-CNN; reported results include 96.2% accuracy, 97% IoU for Mask R-CNN, and successful one-minute tracking in five trials (Guna et al., 2024). In medical-image synthesis, MAMBO becomes the MAMmography ensemBle mOdel, a three-stage patch-based diffusion framework that generates mammograms up to 3840×3840 pixels, achieved a best anomaly-detection IoU of 0.423, and produced synthetic images that radiologists identified with only 0.547 accuracy in a real-versus-synthetic test (Škipina et al., 10 Jun 2025). In medical segmentation, MAMBO-NET denotes Multi-causal Aware Modeling Backdoor-Intervention Optimization, which introduces latent Gaussian modeling of confusion factors and a backdoor-style intervention objective $\mathrm{S/N}>4$5 for medical image segmentation across five datasets (Yu et al., 28 May 2025). In few-shot OOD detection, Mambo refers to a background-prompt-based foreground-background decomposition framework with adaptive patch selection, introduced to improve robustness in FS-OOD and near-OOD settings (Cai et al., 25 Sep 2025).

Taken together, these later usages show that “Mambo” functions as a recurrent acronymic label across disparate research programs. In practice, disambiguation depends entirely on neighboring terms: in astronomy it usually denotes the IRAM millimeter bolometer, in materials semantics an ontology, in optimization a Bayesian algorithm, in systems a DBI framework, and in recent machine learning a growing set of unrelated model names (Aravena et al., 2010, Piane et al., 2024, Wang et al., 2022, Wichelmann et al., 2023).