Search

Travel planning is a challenging and time-consuming task that aims to find an itinerary which satisfies multiple, interdependent constraints regarding flights, accommodations, attractions, and other travel arrangements. In this paper, we propose To the Globe (TTG), a real-time demo system that takes natural language requests from users, translates it to symbolic form via a fine-tuned Large Language Model, and produces optimal travel itineraries with Mixed Integer Linear Programming solvers. The overall system takes ~5 seconds to reply to the user request with guaranteed itineraries. To train TTG, we develop a synthetic data pipeline that generates user requests, flight and hotel information in symbolic form without human annotations, based on the statistics of real-world datasets, and fine-tune an LLM to translate NL user requests to their symbolic form, which is sent to the symbolic solver to compute optimal itineraries. Our NL-symbolic translation achieves ~91% exact match in a backtranslation metric (i.e., whether the estimated symbolic form of generated natural language matches the groundtruth), and its returned itineraries have a ratio of 0.979 compared to the optimal cost of the ground truth user request. When evaluated by users, TTG achieves consistently high Net Promoter Scores (NPS) of 35-40% on generated itinerary.

We numerically investigate the effect of coupling a two-dimensional many-body localized system to a finite heat bath, using shallow quantum circuits as a variational ansatz. Specifically, we simulate optical lattice experiments with two components of ultracold bosons, where only one species is subject to a random disorder potential and the other acts as a heat bath. We obtain a filling fraction dependent phase diagram with a critical filling consistent with experiments. We also calculate two-point correlation functions and the quantum mutual information between sites. We observe a distribution of the quantum mutual information in the many-body localized regime which is consistent with the presence of resonances, similar to those observed in one-dimensional many-body localized systems., Comment: 8 pages, 5 figures

Tokenization is associated with many poorly understood shortcomings in language models (LMs), yet remains an important component for long sequence scaling purposes. This work studies how tokenization impacts model performance by analyzing and comparing the stochastic behavior of tokenized models with their byte-level, or token-free, counterparts. We discover that, even when the two models are statistically equivalent, their predictive distributions over the next byte can be substantially different, a phenomenon we term as "tokenization bias''. To fully characterize this phenomenon, we introduce the Byte-Token Representation Lemma, a framework that establishes a mapping between the learned token distribution and its equivalent byte-level distribution. From this result, we develop a next-byte sampling algorithm that eliminates tokenization bias without requiring further training or optimization. In other words, this enables zero-shot conversion of tokenized LMs into statistically equivalent token-free ones. We demonstrate its broad applicability with two use cases: fill-in-the-middle (FIM) tasks and model ensembles. In FIM tasks where input prompts may terminate mid-token, leading to out-of-distribution tokenization, our method mitigates performance degradation and achieves an approximately 18% improvement in FIM coding benchmarks, consistently outperforming the standard token healing fix. For model ensembles where each model employs a distinct vocabulary, our approach enables seamless integration, resulting in improved performance (up to 3.7%) over individual models across various standard baselines in reasoning, knowledge, and coding.

Background. In the last decades, several life science resources have structured data using the same framework and made these accessible using the same query language to facilitate interoperability. Knowledge graphs have seen increased adoption in bioinformatics due to their advantages for representing data in a generic graph format. For example, yummydata.org catalogs more than 60 knowledge graphs accessible through SPARQL, a technical query language. Although SPARQL allows powerful, expressive queries, even across physically distributed knowledge graphs, formulating such queries is a challenge for most users. Therefore, to guide users in retrieving the relevant data, many of these resources provide representative examples. These examples can also be an important source of information for machine learning, if a sufficiently large number of examples are provided and published in a common, machine-readable and standardized format across different resources. Findings. We introduce a large collection of human-written natural language questions and their corresponding SPARQL queries over federated bioinformatics knowledge graphs (KGs) collected for several years across different research groups at the SIB Swiss Institute of Bioinformatics. The collection comprises more than 1000 example questions and queries, including 65 federated queries. We propose a methodology to uniformly represent the examples with minimal metadata, based on existing standards. Furthermore, we introduce an extensive set of open-source applications, including query graph visualizations and smart query editors, easily reusable by KG maintainers who adopt the proposed methodology. Conclusions. We encourage the community to adopt and extend the proposed methodology, towards richer KG metadata and improved Semantic Web services.

We prove that the complement of a very general pair of hypersurfaces of total degree $2n$ in $\mathbb{P}^n$ is algebraically hyperbolic modulo a proper closed subvariety. This provides evidence towards conjectures of Lang-Vojta and Green-Griffiths, and partially extends previous work of Chen, Pacienza-Rousseau, and Chen-Riedl and the third author., Comment: 14 pages, comments welcome!

Renormalization of physical quantities for quantum field theories in curved spacetimes can be achieved via the consistent subtraction of counterterms within a regularization scheme such as a point-splitting method. Pragmatic mode-sum regularization (PMR) is a point-splitting method which is particularly suitable for rotating black hole spacetimes. We extend and tailor the t-splitting variant of PMR specifically for the interior of a Kerr black hole on the axis of rotation, focusing on a minimally-coupled massless scalar field in the physically-motivated Unruh state. The method addresses unique challenges in the black hole interior that do not occur outside. In particular, while the infinite sum over multipolar number l converges in the black hole exterior, it diverges in the interior, necessitating the subtraction of a so-called intermediate divergence which includes introducing an additional "small" split in the direction of the polar angle. This procedure is outlined and justified, along with the standard PMR method's counterterms subtraction. We apply this method to calculate the renormalized energy-momentum fluxes $\langle T_{uu}\rangle^U_\text{ren}$, $\langle T_{vv}\rangle^U_\text{ren}$ (where u and v are the standard Eddington coordinates) and the renormalized field square $\langle \Phi^2\rangle^U_\text{ren}$ throughout the black hole interior, spanning from (just off) the event horizon to (just off) the inner horizon. Special emphasis is placed on the inner horizon vicinity, where our t-splitting results for the fluxes asymptote to those obtained directly at the inner horizon using a different method in a previous work. In an Appendix, we develop an alternative t-splitting PMR variant which does not include the intermediate divergence subtraction. We utilize it for independent computations that are used to verify the standard t-splitting variant presented in the main text., Comment: 57 pages, 19 figures. Including a supplemental Mathematica notebook

Digital phenotyping offers a novel and cost-efficient approach for managing depression and anxiety. Previous studies, often limited to small-to-medium or specific populations, may lack generalizability. We conducted a cross-sectional analysis of data from 10,129 participants recruited from a UK-based general population between June 2020 and August 2022. Participants shared wearable (Fitbit) data and self-reported questionnaires on depression (PHQ-8), anxiety (GAD-7), and mood via a study app. We first examined the correlations between PHQ-8/GAD-7 scores and wearable-derived features, demographics, health data, and mood assessments. Subsequently, unsupervised clustering was used to identify behavioural patterns associated with depression or anxiety. Finally, we employed separate XGBoost models to predict depression and anxiety and compared the results using different subsets of features. We observed significant associations between the severity of depression and anxiety with several factors, including mood, age, gender, BMI, sleep patterns, physical activity, and heart rate. Clustering analysis revealed that participants simultaneously exhibiting lower physical activity levels and higher heart rates reported more severe symptoms. Prediction models incorporating all types of variables achieved the best performance ($R^2$=0.41, MAE=3.42 for depression; $R^2$=0.31, MAE=3.50 for anxiety) compared to those using subsets of variables. This study identified potential indicators for depression and anxiety, highlighting the utility of digital phenotyping and machine learning technologies for rapid screening of mental disorders in general populations. These findings provide robust real-world insights for future healthcare applications.

The ramp reversal memory (RRM) is a non-volatile memory effect previously observed in correlated oxides exhibiting temperature-driven metal-insulator transitions (MITs). In essence, when a system displaying RRM is heated to a specific temperature within the MIT regime - where metallic and insulating domains coexist - and then cooled by reversing the temperature ramp, the resistance increases in the subsequent heating cycle. Crucially, this increase occurs only in the vicinity of the reversal temperature, indicating that the system 'remembers' this temperature. However, this memory is erased in the next heating loop. While such an effect could potentially manifest in various systems, to date, it has only been reported in thin films of correlated transition metal oxides, including VO2, V2O3, and NdNiO3. In this work, we report the observation of RRM in macroscopic crystals of the layered material 1T-TaS2, which undergoes an MIT near 190 K along charge-density wave transitions. Our findings provide compelling evidence that RRM is a general phenomenon, extending beyond the previously studied oxides. Interestingly, the RRM in TaS2 displays significantly different characteristics: it is observed when reversing from cooling to heating (as opposed to heating to cooling), and its magnitude - representing the 'strength' of the memory - is nearly an order of magnitude larger than in correlated oxides. While we discuss potential mechanisms for the RRM in TaS2, a comprehensive first-principles model is still lacking. We hope that this study will prompt further investigation into the underlying mechanisms of ramp reversal memory, enhancing our understanding of this intriguing phenomenon., Comment: A supporting information file follows the main article

Training neural network classifiers on datasets contaminated with noisy labels significantly increases the risk of overfitting. Thus, effectively implementing Early Stopping in noisy label environments is crucial. Under ideal circumstances, Early Stopping utilises a validation set uncorrupted by label noise to effectively monitor generalisation during training. However, obtaining a noise-free validation dataset can be costly and challenging to obtain. This study establishes that, in many typical learning environments, a noise-free validation set is not necessary for effective Early Stopping. Instead, near-optimal results can be achieved by monitoring accuracy on a noisy dataset - drawn from the same distribution as the noisy training set. Referred to as `Noisy Early Stopping' (NES), this method simplifies and reduces the cost of implementing Early Stopping. We provide theoretical insights into the conditions under which this method is effective and empirically demonstrate its robust performance across standard benchmarks using common loss functions.

Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles. Learning the dynamics of such systems is essential for predicting the temporal evolution of populations across novel samples and unseen environments. Flow-based models allow for learning these dynamics at the population level - they model the evolution of the entire distribution of samples. However, current flow-based models are limited to a single initial population and a set of predefined conditions which describe different dynamics. We argue that multiple processes in natural sciences have to be represented as vector fields on the Wasserstein manifold of probability densities. That is, the change of the population at any moment in time depends on the population itself due to the interactions between samples. In particular, this is crucial for personalized medicine where the development of diseases and their respective treatment response depends on the microenvironment of cells specific to each patient. We propose Meta Flow Matching (MFM), a practical approach to integrating along these vector fields on the Wasserstein manifold by amortizing the flow model over the initial populations. Namely, we embed the population of samples using a Graph Neural Network (GNN) and use these embeddings to train a Flow Matching model. This gives MFM the ability to generalize over the initial distributions unlike previously proposed methods. We demonstrate the ability of MFM to improve prediction of individual treatment responses on a large scale multi-patient single-cell drug screen dataset.

The accuracy of mapping agricultural fields across large areas is steadily improving with high-resolution satellite imagery and deep learning (DL) models, even in regions where fields are small and geometrically irregular. However, developing effective DL models often requires large, expensive label datasets, typically available only for specific years or locations. This limits the ability to create annual maps essential for agricultural monitoring, as domain shifts occur between years and regions due to changes in farming practices and environmental conditions. The challenge is to design a model flexible enough to account for these shifts without needing yearly labels. While domain adaptation techniques or semi-supervised training are common solutions, we explored enhancing the model's generalization power. Our results indicate that a holistic approach is essential, combining methods to improve generalization. Specifically, using an area-based loss function, such as Tversky-focal loss (TFL), significantly improved predictions across multiple years. The use of different augmentation techniques helped to encode different types of invariance, particularly photometric augmentations encoded invariance to brightness changes, though they increased false positives. The combination of photometric augmentation, TFL loss, and MC-dropout produced the best results, although dropout alone led to more false negatives in subsequent year predictions. Additionally, the choice of input normalization had a significant impact, with the best results obtained when statistics were calculated either locally or across the entire dataset over all bands (lab and gab). We developed a workflow that enabled a U-Net model to generate effective multi-year crop maps over large areas. Our code, available at: https://github.com/agroimpacts/cnn-generalization-enhancement, will be regularly updated with improvements.

We study the supersymmetric structure of the spin current in four dimensional $\mathcal{N}=1$ supersymmetric theories. By coupling the stress tensor multiplet to a vierbein multiplet we identify a spin current with the Hodge dual of the bottom component of the stress tensor multiplet in a wide range of theories. This implies that in holographic theories the Hodge dual of the $R$ current may serve as a spin current, paving the way for holographic studies of theories with background torsion.

Serverless computing is a growing and maturing field that is the focus of much research, industry interest and adoption. Previous works exploring Functions-as-a-Service providers have focused primarily on the most well known providers AWS Lambda, Google Cloud Functions and Microsoft Azure Functions without exploring other providers in similar detail. In this work, we conduct the first detailed review of ten currently publicly available FaaS platforms exploring everything from their history, to their features and pricing to where they sit within the overall public FaaS landscape, before making a number of observations as to the state of the FaaS., Comment: This paper has been accepted to the 17th IEEE International Conference on Cloud Computing (CLOUD). This paper contains 10 pages and 2 figures

As a potential non-invasive biomarker for ischaemic stroke, intracranial arterial calcification (IAC) could be used for stroke risk assessment on CT head scans routinely acquired for other reasons (e.g. trauma, confusion). Artificial intelligence methods can support IAC scoring, but they have not yet been developed for clinical imaging. Large heterogeneous clinical CT datasets are necessary for the training of such methods, but they exhibit expected and unexpected data anomalies. Using CTs from a large clinical trial, the third International Stroke Trial (IST-3), we propose a pipeline that uses as input non-enhanced CT scans to output regions of interest capturing selected large intracranial arteries for IAC scoring. Our method uses co-registration with templates. We focus on quality control, using information presence along the z-axis of the imaging to group and apply similarity measures (structural similarity index measure) to triage assessment of individual image series. Additionally, we propose superimposing thresholded binary masks of the series to inspect large quantities of data in parallel. We identify and exclude unrecoverable samples and registration failures. In total, our pipeline processes 10,659 CT series, rejecting 4,322 (41%) in the entire process, 1,450 (14% of the total) during quality control, and outputting 6,337 series. Our pipeline enables effective and efficient region of interest localisation for targeted IAC segmentation., Comment: Accepted at the 2nd Data Engineering in Medical Imaging workshop @ MICCAI 2024

Large language models (LLMs) with billions of parameters excel at predicting the next token in a sequence. Recent work computes non-vacuous compression-based generalization bounds for LLMs, but these bounds are vacuous for large models at the billion-parameter scale. Moreover, these bounds are obtained through restrictive compression techniques, bounding compressed models that generate low-quality text. Additionally, the tightness of these existing bounds depends on the number of IID documents in a training set rather than the much larger number of non-IID constituent tokens, leaving untapped potential for tighter bounds. In this work, we instead use properties of martingales to derive generalization bounds that benefit from the vast number of tokens in LLM training sets. Since a dataset contains far more tokens than documents, our generalization bounds not only tolerate but actually benefit from far less restrictive compression schemes. With Monarch matrices, Kronecker factorizations, and post-training quantization, we achieve non-vacuous generalization bounds for LLMs as large as LLaMA2-70B. Unlike previous approaches, our work achieves the first non-vacuous bounds for models that are deployed in practice and generate high-quality text.

It is challenging to scale time series forecasting models such that they forecast accurately for multiple distinct domains and datasets, all with potentially different underlying collection procedures (e.g., sample resolution), patterns (e.g., periodicity), and prediction requirements (e.g., reconstruction vs. forecasting). We call this general task universal forecasting. Existing methods usually assume that input data is regularly sampled, and they forecast to pre-determined horizons, resulting in failure to generalise outside of the scope of their training. We propose the DAM - a neural model that takes randomly sampled histories and outputs an adjustable basis composition as a continuous function of time for forecasting to non-fixed horizons. It involves three key components: (1) a flexible approach for using randomly sampled histories from a long-tail distribution, that enables an efficient global perspective of the underlying temporal dynamics while retaining focus on the recent history; (2) a transformer backbone that is trained on these actively sampled histories to produce, as representational output, (3) the basis coefficients of a continuous function of time. We show that a single univariate DAM, trained on 25 time series datasets, either outperformed or closely matched existing SoTA models at multivariate long-term forecasting across 18 datasets, including 8 held-out for zero-shot transfer, even though these models were trained to specialise for each dataset-horizon combination. This single DAM excels at zero-shot transfer and very-long-term forecasting, performs well at imputation, is interpretable via basis function composition and attention, can be tuned for different inference-cost requirements, is robust to missing and irregularly sampled data {by design}.

Gravitational waves (GWs) passing through a viscous shell of matter are expected to be damped resulting in an increase in the temperature of the fluid as energy is transferred to it from the GWs. In previous work we constructed a model for this process, obtaining an expression for the temperature distribution inside the shell, and it was shown that the temperature increase can be astrophysically significant. In this paper we extend the analysis to GW heating and damping following a binary neutron star merger, GW heating during a core-collapse supernova, and primordial gravitational waves.

ULTRASAT (ULtra-violet TRansient Astronomy SATellite) is a wide-angle space telescope that will perform a deep time-resolved all-sky survey in the near-ultraviolet (NUV) spectrum. The science objectives are the detection of counterparts to short-lived transient astronomical events such as gravitational wave sources and supernovae. The mission is led by the Weizmann Institute of Science and is planned for launch in 2026 in collaboration with the Israeli Space Agency and NASA. DESY will provide the UV camera, composed by the detector assembly located in the telescope focal plane and the remote electronics unit. The camera is composed out of four back-metallized CMOS Image Sensors (CIS) manufactured in the 4T, dual gain Tower process. As part of the radiation qualification of the camera, Single Event Effect (SEE) testing has been performed by irradiating the sensor with heavy ions at the RADEF, Jyvaskyla facility. Preliminary results of both Single Event Upset (SEU) and Single Event Latch-up (SEL) occurrence rate in the sensor are presented. Additionally, an in-orbit SEE rate simulation has been performed in order to gain preliminary knowledge about the expected effect of SEE on the mission.

Consistent hashing is employed in distributed systems and networking applications to evenly and effectively distribute data across a cluster of nodes. This paper introduces BinomialHash, a consistent hashing algorithm that operates in constant time and requires minimal memory. We provide a detailed explanation of the algorithm, offer a pseudo-code implementation, and formally establish its strong theoretical guarantees.

We study integral plane curves meeting at a single unibranch point and show that such curves must satisfy two equivalent conditions. A numeric condition: the local invariants of the curves at the contact point must be arithmetically related. A geometric condition: the tropical curves that we associate to the contact point must be isomorphic. Moreover, we prove closed formulas for the delta-invariant of a unibranch singularity, and for the dimension of the loci of curves with an assigned unibranch point. Our work is motivated by interest in the Lang exceptional set., Comment: 23 pages, comments welcome!

We present a lattice model for polymer solutions, explicitly incorporating interactions with solvent molecules and the contribution of vacancies. By exploiting the well-known analogy between polymer systems and the $O(n)$-vector spin model in the limit $n \to 0$, we derive an exact field-theoretic expression for the partition function of the system. The latter is then evaluated at the saddle-point, providing a mean-field estimate of the free energy. The resulting expression, which conforms to the Flory-Huggins type, is then used to analyze the system's stability with respect to phase separation, complemented by a numerical approach based on convex hull evaluation. We demonstrate that this simple lattice model can effectively explain the behavior of polymer systems in explicit solvent, which has been predominantly investigated through numerical simulations. This includes both single-chain systems and polymer solutions. Our findings emphasize the fundamental role of vacancies whose presence, rendering the system a ternary mixture, is a crucial factor in the observed phase behavior., Comment: 7 pages, 4 figures (main); 11 pages, 4 figures (suppl info); submitted for publication

Recent work has demonstrated both benefits and limitations from using supervised approaches (without temporal-difference learning) for offline reinforcement learning. While off-policy reinforcement learning provides a promising approach for improving performance beyond supervised approaches, we observe that training is often inefficient and unstable due to temporal difference bootstrapping. In this paper we propose a best-of-both approach by first learning the behavior policy and critic with supervised learning, before improving with off-policy reinforcement learning. Specifically, we demonstrate improved efficiency by pre-training with a supervised Monte-Carlo value-error, making use of commonly neglected downstream information from the provided offline trajectories. We find that we are able to more than halve the training time of the considered offline algorithms on standard benchmarks, and surprisingly also achieve greater stability. We further build on the importance of having consistent policy and value functions to propose novel hybrid algorithms, TD3+BC+CQL and EDAC+BC, that regularize both the actor and the critic towards the behavior policy. This helps to more reliably improve on the behavior policy when learning from limited human demonstrations. Code is available at https://github.com/AdamJelley/EfficientOfflineRL

Realising the affect of composition and processing condition on bainite microstructures is often challenging, owing to the intricate distribution of the constituent phases. In this work, scanning electron micrographs of non-isothermally transformed bainite, with martensite-austenite (MA) islands, are analysed to relate the microstructures to the composition and quench-stop temperature. The inadequacy of the MA-islands' geometric features, namely aspect ratio, polygon area and compactness, in establishing this relation is made evident from Kullback-Leibler (KL) divergence at the outset. Clustering the bainite microstructures, following a combination of feature extraction and dimensionality reduction, further fails to realise the affect of composition and processing temperature. Deep-learning analysis of the individual MA islands, in contrast to the bainite microstructures, yields interpretable clusters with characteristically distinct size and morphology. These five clusters, referred to as fine- and coarse-dendrite, fine- and coarse-polygon and elongated, are exceptionally discernible and can be adopted to describe any MA island. Characterising the bainite microstructures, based on the distribution of the interpretable MA-island clusters, generates \textit{fingerprints} that sufficiently relates the composition and processing conditions with the microstructures.

Rate of grain growth, which aides in achieving desired properties in polycrystalline materials, is conventionally estimated by measuring the size of grains and tracking its change in micrographs reflecting the temporal evolution. Techniques adopting this conventional approach demand an absolute distinction between the grains and the interface separating them to yield an accurate result. Edge-detection, segmentation and other deep-learning algorithms are increasingly adopted to expose the network of boundaries and the associated grains precisely. An alternate approach for measuring grain-growth kinetics, that curtails the need for advanced image-processing treatment, is presented in this work. Grain-growth rate in the current technique is ascertained by \textit{counting} the number of triple-( and quadruple-) junctions, and monitoring its change during the microstructural evolution. The shifted focus of this junction-based treatment minimises the significance of a well-defined grain-boundary network, and consequently, the involvement of the sophisticated techniques that expose them. A regression-based object-detection algorithm is extended to realise, and count, the number of junctions in polycrystalline microstructures. By examining the change in the number of junctions with time, the growth rate is subsequently determined.Growth kinetics estimated by the present junction-based approach, across a wide-range of multiphase polycrystalline microstructures, agree convincingly with the outcomes of the conventional treatment.Besides offering a novel technique for grain-growth measurement, the analysis accompanying the current work unravels a trend, compatible with the topological events, in the progressive evolution of the triple-junctions count. The present approach, through its underlying algorithm, provides a promising option for monitoring grain-growth during in-situ investigations.

Achieving desired properties in bainite steels with MA islands demands understanding the affect of processing conditions and composition on their size and morphology. Generally, this understanding is gained by studying the change in the size and morphology of the MA islands with composition and the processing conditions. In the present work, around 8500 MA islands dispersed across of approximately 1500 bainite microstructures are investigated to comprehend the influence of composition and heat treatment cycle on the geometric features. The geometric features considered in this study include polygon area metric, compactness and aspect ratio. A thorough statistical analysis of these features across bainite steels of different compositions and processing conditions unravel that, though there are minor changes, no characteristic variation is introduced in the size and morphology of the MA islands. In other words, the distribution of the various forms of MA islands are almost identical in bainite steels of different chemistry and heat treatment, thereby indicating the inadequacy of geometric features in explicating the affect of processing conditions on microstructure.

Buildings play a crucial role in human well-being, influencing occupant comfort, health, and safety. Additionally, they contribute significantly to global energy consumption, accounting for one-third of total energy usage, and carbon emissions. Optimizing building performance presents a vital opportunity to combat climate change and promote human flourishing. However, research in building analytics has been hampered by the lack of accessible, available, and comprehensive real-world datasets on multiple building operations. In this paper, we introduce the Building TimeSeries (BTS) dataset. Our dataset covers three buildings over a three-year period, comprising more than ten thousand timeseries data points with hundreds of unique ontologies. Moreover, the metadata is standardized using the Brick schema. To demonstrate the utility of this dataset, we performed benchmarks on two tasks: timeseries ontology classification and zero-shot forecasting. These tasks represent an essential initial step in addressing challenges related to interoperability in building analytics. Access to the dataset and the code used for benchmarking are available here: https://github.com/cruiseresearchgroup/DIEF_BTS ., Comment: 21 pages, 2 figures, 9 tables, under review

Bacterial communities are pivotal to maintaining ecological function and preserving the rich tapestry of biological diversity. The rapid development of environmental sequencing technologies, such as metagenomics, has revolutionized our capacity to probe such diversity. However, despite these advances, a theoretical understanding connecting empirical data with ecosystem modelling, in particular in the framework of disordered systems akin to spin glasses, is still in its infancy. Here, we present a comprehensive framework using theories of disordered systems to decode microbiome data, which offers insight into the ecological forces that shape macroecological states. By employing the quenched disordered generalized Lotka-Volterra model, we analyze species abundance data in healthy and diseased human gut microbiomes. Results reveal the emergence of two distinct patterns of species-interaction networks, elucidating the pathways through which dysbiosis may drive microbiome instability. Interaction patterns thus provide a window into the systemic shifts accompanying the transition from health to disease, offering a new perspective on the dynamics of the microbial community. Our findings suggest the potential of disordered systems theory to characterize microbiomes by capturing the essence of ecological interactions and their consequences on stability and functioning, leveraging our understanding of the linkages of dysbiosis and microbial dynamics., Comment: 10 pages, 4 figures