Search

We describe a family of architectures to support transductive inference by allowing memory to grow to a finite but a-priori unknown bound while making efficient use of finite resources for inference. Current architectures use such resources to represent data either eidetically over a finite span ("context" in Transformers), or fading over an infinite span (in State Space Models, or SSMs). Recent hybrid architectures have combined eidetic and fading memory, but with limitations that do not allow the designer or the learning process to seamlessly modulate the two, nor to extend the eidetic memory span. We leverage ideas from Stochastic Realization Theory to develop a class of models called B'MOJO to seamlessly combine eidetic and fading memory within an elementary composable module. The overall architecture can be used to implement models that can access short-term eidetic memory "in-context," permanent structural memory "in-weights," fading memory "in-state," and long-term eidetic memory "in-storage" by natively incorporating retrieval from an asynchronously updated memory. We show that Transformers, existing SSMs such as Mamba, and hybrid architectures such as Jamba are special cases of B'MOJO and describe a basic implementation, to be open sourced, that can be stacked and scaled efficiently in hardware. We test B'MOJO on transductive inference tasks, such as associative recall, where it outperforms existing SSMs and Hybrid models; as a baseline, we test ordinary language modeling where B'MOJO achieves perplexity comparable to similarly-sized Transformers and SSMs up to 1.4B parameters, while being up to 10% faster to train. Finally, we show that B'MOJO's ability to modulate eidetic and fading memory results in better inference on longer sequences tested up to 32K tokens, four-fold the length of the longest sequences seen during training.

As the closest transiting hot Jupiter to Earth, HD 189733b has been the benchmark planet for atmospheric characterization. It has also been the anchor point for much of our theoretical understanding of exoplanet atmospheres from composition, chemistry, aerosols to atmospheric dynamics, escape, and modeling techniques. Prior studies of HD 189733b have detected carbon and oxygen-bearing molecules H2O and CO in the atmosphere. The presence of CO2 and CH4 has been claimed but later disputed. The inferred metallicity based on these measurements, a key parameter in tracing planet formation locations, varies from depletion to enhancement, hindered by limited wavelength coverage and precision of the observations. Here we report detections of H2O (13.4 sigma), CO2 (11.2 sigma), CO (5 sigma), and H2S (4.5 sigma) in the transmission spectrum (2.4-5 micron) of HD 189733b. With an equilibrium temperature of ~1200K, H2O, CO, and H2S are the main reservoirs for oxygen, carbon, and sulfur. Based on the measured abundances of these three major volatile elements, we infer an atmospheric metallicity of 3-5 times stellar. The upper limit on the methane abundance at 5 sigma is 0.1 ppm which indicates a low carbon-to-oxygen ratio (<0.2), suggesting formation through the accretion of water-rich icy planetesimals. The low oxygen-to-sulfur and carbon-to-sulfur ratios also support the planetesimal accretion formation pathway., Comment: Published online in Nature on July 8th, 2024

Self-attention performs well in long context but has quadratic complexity. Existing RNN layers have linear complexity, but their performance in long context is limited by the expressive power of their hidden state. We propose a new class of sequence modeling layers with linear complexity and an expressive hidden state. The key idea is to make the hidden state a machine learning model itself, and the update rule a step of self-supervised learning. Since the hidden state is updated by training even on test sequences, our layers are called Test-Time Training (TTT) layers. We consider two instantiations: TTT-Linear and TTT-MLP, whose hidden state is a linear model and a two-layer MLP respectively. We evaluate our instantiations at the scale of 125M to 1.3B parameters, comparing with a strong Transformer and Mamba, a modern RNN. Both TTT-Linear and TTT-MLP match or exceed the baselines. Similar to Transformer, they can keep reducing perplexity by conditioning on more tokens, while Mamba cannot after 16k context. With preliminary systems optimization, TTT-Linear is already faster than Transformer at 8k context and matches Mamba in wall-clock time. TTT-MLP still faces challenges in memory I/O, but shows larger potential in long context, pointing to a promising direction for future research.

We infer the growth of large scale structure over the redshift range $0.4\lesssim z \lesssim 1$ from the cross-correlation of spectroscopically calibrated Luminous Red Galaxies (LRGs) selected from the Dark Energy Spectroscopic Instrument (DESI) legacy imaging survey with CMB lensing maps reconstructed from the latest Planck and ACT data. We adopt a hybrid effective field theory (HEFT) model that robustly regulates the cosmological information obtainable from smaller scales, such that our cosmological constraints are reliably derived from the (predominantly) linear regime. We perform an extensive set of bandpower- and parameter-level systematics checks to ensure the robustness of our results and to characterize the uniformity of the LRG sample. We demonstrate that our results are stable to a wide range of modeling assumptions, finding excellent agreement with a linear theory analysis performed on a restricted range of scales. From a tomographic analysis of the four LRG photometric redshift bins we find that the rate of structure growth is consistent with $\Lambda$CDM with an overall amplitude that is $\simeq5-7\%$ lower than predicted by primary CMB measurements with modest $(\sim2\sigma)$ statistical significance. From the combined analysis of all four bins and their cross-correlations with Planck we obtain $S_8 = 0.765\pm0.023$, which is less discrepant with primary CMB measurements than previous DESI LRG cross Planck CMB lensing results. From the cross-correlation with ACT we obtain $S_8 = 0.790^{+0.024}_{-0.027}$, while when jointly analyzing Planck and ACT we find $S_8 = 0.775^{+0.019}_{-0.022}$ from our data alone and $\sigma_8 = 0.772^{+0.020}_{-0.023}$ with the addition of BAO data. These constraints are consistent with the latest Planck primary CMB analyses at the $\simeq 1.6-2.2\sigma$ level, and are in excellent agreement with galaxy lensing surveys., Comment: 60 pages, 26 figures, comments welcome

We present a high-significance cross-correlation of CMB lensing maps from the Atacama Cosmology Telescope (ACT) Data Release 6 (DR6) with spectroscopically calibrated luminous red galaxies (LRGs) from the Dark Energy Spectroscopic Instrument (DESI). We detect this cross-correlation at a significance of 38$\sigma$; combining our measurement with the Planck Public Release 4 (PR4) lensing map, we detect the cross-correlation at 50$\sigma$. Fitting this jointly with the galaxy auto-correlation power spectrum to break the galaxy bias degeneracy with $\sigma_8$, we perform a tomographic analysis in four LRG redshift bins spanning $0.4 \le z \le 1.0$ to constrain the amplitude of matter density fluctuations through the parameter combination $S_8^\times = \sigma_8 \left(\Omega_m / 0.3\right)^{0.4}$. Prior to unblinding, we confirm with extragalactic simulations that foreground biases are negligible and carry out a comprehensive suite of null and consistency tests. Using a hybrid effective field theory (HEFT) model that allows scales as small as $k_{\rm max}=0.6$ $h/{\rm Mpc}$, we obtain a 3.3% constraint on $S_8^\times = \sigma_8 \left(\Omega_m / 0.3\right)^{0.4} = 0.792^{+0.024}_{-0.028}$ from ACT data, as well as constraints on $S_8^\times(z)$ that probe structure formation over cosmic time. Our result is consistent with the early-universe extrapolation from primary CMB anisotropies measured by Planck PR4 within 1.2$\sigma$. Jointly fitting ACT and Planck lensing cross-correlations we obtain a 2.7% constraint of $S_8^\times = 0.776^{+0.019}_{-0.021}$, which is consistent with the Planck early-universe extrapolation within 2.1$\sigma$, with the lowest redshift bin showing the largest difference in mean. The latter may motivate further CMB lensing tomography analyses at $z<0.6$ to assess the impact of potential systematics or the consistency of the $\Lambda$CDM model over cosmic time., Comment: Prepared for submission to JCAP (47 pages, 13 figures)

Backdoors are hidden behaviors that are only triggered once an AI system has been deployed. Bad actors looking to create successful backdoors must design them to avoid activation during training and evaluation. Since data used in these stages often only contains information about events that have already occurred, a component of a simple backdoor trigger could be a model recognizing data that is in the future relative to when it was trained. Through prompting experiments and by probing internal activations, we show that current large language models (LLMs) can distinguish past from future events, with probes on model activations achieving 90% accuracy. We train models with backdoors triggered by a temporal distributional shift; they activate when the model is exposed to news headlines beyond their training cut-off dates. Fine-tuning on helpful, harmless and honest (HHH) data does not work well for removing simpler backdoor triggers but is effective on our backdoored models, although this distinction is smaller for the larger-scale model we tested. We also find that an activation-steering vector representing a model's internal representation of the date influences the rate of backdoor activation. We take these results as initial evidence that, at least for models at the modest scale we test, standard safety measures are enough to remove these backdoors.

Despite advancements in Autoencoders (AEs) for tasks like dimensionality reduction, representation learning and data generation, they remain vulnerable to adversarial attacks. Variational Autoencoders (VAEs), with their probabilistic approach to disentangling latent spaces, show stronger resistance to such perturbations compared to deterministic AEs; however, their resilience against adversarial inputs is still a concern. This study evaluates the robustness of VAEs against non-targeted adversarial attacks by optimizing minimal sample-specific perturbations to cause maximal damage across diverse demographic subgroups (combinations of age and gender). We investigate two questions: whether there are robustness disparities among subgroups, and what factors contribute to these disparities, such as data scarcity and representation entanglement. Our findings reveal that robustness disparities exist but are not always correlated with the size of the subgroup. By using downstream gender and age classifiers and examining latent embeddings, we highlight the vulnerability of subgroups like older women, who are prone to misclassification due to adversarial perturbations pushing their representations toward those of other subgroups.

The advancements of machine learning-based (ML) decision-making algorithms created various research and industrial opportunities. One of these areas is ML-based near-real-time network management applications (xApps) in Open-Radio Access Network (O-RAN). Normally, xApps are designed solely for the desired objectives, and fine-tuned for deployment. However, telecommunication companies can employ multiple xApps and deploy them in overlapping areas. Consider the different design objectives of xApps, the deployment might cause conflicts. To prevent such conflicts, we proposed the xApp distillation method that distills knowledge from multiple xApps, then uses this knowledge to train a single model that has retained the capabilities of Previous xApps. Performance evaluations show that compared conflict mitigation schemes can cause up to six times more network outages than xApp distillation in some cases., Comment: 5 Pages, 4 figures

With the number of supernovae observed expected to drastically increase thanks to large-scale surveys like the Dark Energy Spectroscopic Instrument (DESI), it is necessary that the tools we use to classify these objects keep up with this increase. We previously created Supernova Tagging and Classification (STag) to address this problem by employing machine learning techniques alongside logistic regression in order to assign 'tags' to spectra based on spectral features. STag II is a continuation of this work, which now makes use of model supernova spectra combined with real DESI spectra in order to train STag to better deal with realistic data. We also make use of the rlap score as a trustworthiness cut, making for a more robust and accurate supernova classifier than before., Comment: 27 pages, 7 figures. The STag code is available at https://github.com/wdavison909/STag

In this paper we examine the limitations of Large Language Models (LLMs) for complex reasoning tasks. Although recent works have started to employ formal languages as an intermediate representation for reasoning tasks, they often face challenges in accurately generating and refining these formal specifications to ensure correctness. To address these issues, this paper proposes Logic-LM++, an improvement on Logic-LM . It uses the ability of LLMs to do pairwise comparisons, allowing the evaluation of the refinements suggested by the LLM. The paper demonstrates that Logic-LM++ outperforms Logic-LM and other contemporary techniques across natural language reasoning tasks on three datasets, FOLIO, ProofWriter and AR-LSAT, with an average improvement of 18.5% on standard prompting, 12.3% on chain of thought prompting and 5% on Logic-LM.

We introduce null contractions of the Poincare and relativistic conformal algebras. The longitudinal null contraction involves writing the algebra in lightcone coordinates and contracting one of the null directions. For the Poincare algebra, this yields two non-overlapping co-dimension one Carroll algebras. The transverse contraction is a limit on the spatial dimensions and yields two non-overlapping co-dimension one Galilean algebras. We find, similar to Susskind's original observation of the non-relativistic case, that the Poincare algebra, written in the lightcone coordinates, naturally contains Carrollian sub-algebras in one lower dimension. The effect of the longitudinal contraction, which essentially focusses on the null direction, is to disentangle the two Carroll algebras that now correspond to the symmetries of the two null boundaries. The transverse contraction similarly separates the overlapping Galilean sub-algebras of the original Poincare algebra. We discuss aspects of the conformal case, where we get lower dimensional Carroll Conformal algebras and Schrodinger algebras., Comment: 23 pages

In this work, we propose a two-phased approach for real-time detection and deterrence of ransomware. To achieve this, we leverage the capabilities of eBPF (Extended Berkeley Packet Filter) and artificial intelligence to develop both proactive and reactive methods. In the first phase, we utilize signature based detection, where we employ custom eBPF programs to trace the execution of new processes and perform hash-based analysis against a known ransomware dataset. In the second, we employ a behavior-based technique that focuses on monitoring the process activities using a custom eBPF program and the creation of ransom notes, a prominent indicator of ransomware activity through the use of Natural Language Processing (NLP). By leveraging low-level tracing capabilities of eBPF and integrating NLP based machine learning algorithms, our solution achieves an impressive 99.76% accuracy in identifying ransomware incidents within a few seconds on the onset of zero-day attacks., Comment: 7 pages

LLMs offer valuable capabilities, yet they can be utilized by malicious users to disseminate deceptive information and generate fake news. The growing prevalence of LLMs poses difficulties in crafting detection approaches that remain effective across various text domains. Additionally, the absence of precautionary measures for AI-generated news on online social platforms is concerning. Therefore, there is an urgent need to improve people's ability to differentiate between news articles written by humans and those produced by LLMs. By providing cues in human-written and LLM-generated news, we can help individuals increase their skepticism towards fake LLM-generated news. This paper aims to elucidate simple markers that help individuals distinguish between articles penned by humans and those created by LLMs. To achieve this, we initially collected a dataset comprising 39k news articles authored by humans or generated by four distinct LLMs with varying degrees of fake. We then devise a metric named Entropy-Shift Authorship Signature (ESAS) based on the information theory and entropy principles. The proposed ESAS ranks terms or entities, like POS tagging, within news articles based on their relevance in discerning article authorship. We demonstrate the effectiveness of our metric by showing the high accuracy attained by a basic method, i.e., TF-IDF combined with logistic regression classifier, using a small set of terms with the highest ESAS score. Consequently, we introduce and scrutinize these top ESAS-ranked terms to aid individuals in strengthening their skepticism towards LLM-generated fake news.

Bioelectronic implants with soft mechanics, biocompatibility, and excellent electrical performance enable biomedical implants to record electrophysiological signals and execute interventions within internal organs, promising to revolutionize the diagnosing, monitoring, and treatment of various pathological conditions. However, challenges remain in improving excessive impedance at the bioelectronic-tissue interface and thus the efficacy of electrophysiological signaling and intervention. Here, we devise orbit symmetry breaking in MXene (a low-cost scalability, biocompatible, and conductive 2D layered material, that we refer to as OBXene), that exhibits low bioelectronic-tissue impedance, originating from the out-of-plane charge transfer. Furthermore, the Schottky-induced piezoelectricity stemming from the asymmetric orbital configuration of OBXene facilitates interlayered charge transport in the device. In this study, we report an OBXene-based cardiac patch applied on the left ventricular epicardium of both rodent and porcine models to enable spatiotemporal epicardium mapping and pacing, while coupling the wireless and battery-free operation for long-term real-time recording and closed-loop stimulation.

Magnetic fields connect an array of planetary processes, from atmospheric escape to interior convection. Despite their importance, exoplanet magnetic fields are largely unconstrained by both theory and observation. In this Letter, we propose a novel method for constraining the B field strength of hot gas giants: comparing the velocities of heavy ions and neutral gas with high-resolution spectroscopy. The core concept of this method is that ions are directly deflected by magnetic fields. While neutrals are also affected by B fields via friction with field-accelerated ions, ionic gas should be more strongly coupled to the underlying magnetic field than bulk neutral flow. Hence, measuring the difference between the two velocities yields rough constraints on the B field, provided an estimate of the stellar UV flux is known. We demonstrate that heavy ions are particularly well suited for this technique, because they are less likely to be entrained in complex hydrodynamic outflows than their lighter counterparts. We perform a proof-of-concept calculation with Ba II, an ion whose velocity has been repeatedly measured at high confidence with high-resolution spectroscopy. Our work shows that a 10G magnetic field would produce ~ km/s ion--neutral velocity differences at a microbar, whereas a 50G magnetic field would produce ~20km/s velocity difference. With new leverage on magnetic fields, we will be able to investigate magnetic field generation in the extreme edge cases of hot gas giants, with wide-ranging consequences for planetary interior structure, dynamo theory, and habitability., Comment: Accepted at ApJ Letters. 12 pages; 4 figures; 2 tables

Despite the large body of work on fairness-aware learning for individual modalities like tabular data, images, and text, less work has been done on multimodal data, which fuses various modalities for a comprehensive analysis. In this work, we investigate the fairness and bias implications of multimodal fusion techniques in the context of multimodal AI-based recruitment systems using the FairCVdb dataset. Our results show that early-fusion closely matches the ground truth for both demographics, achieving the lowest MAEs by integrating each modality's unique characteristics. In contrast, late-fusion leads to highly generalized mean scores and higher MAEs. Our findings emphasise the significant potential of early-fusion for accurate and fair applications, even in the presence of demographic biases, compared to late-fusion. Future research could explore alternative fusion strategies and incorporate modality-related fairness constraints to improve fairness. For code and additional insights, visit: https://github.com/Swati17293/Multimodal-AI-Based-Recruitment-FairCVdb

Recent improvements in natural language processing (NLP) and machine learning (ML) and increased mainstream adoption have led to researchers frequently discussing the "democratization" of artificial intelligence. In this paper, we seek to clarify how democratization is understood in NLP and ML publications, through large-scale mixed-methods analyses of papers using the keyword "democra*" published in NLP and adjacent venues. We find that democratization is most frequently used to convey (ease of) access to or use of technologies, without meaningfully engaging with theories of democratization, while research using other invocations of "democra*" tends to be grounded in theories of deliberation and debate. Based on our findings, we call for researchers to enrich their use of the term democratization with appropriate theory, towards democratic technologies beyond superficial access., Comment: EMNLP 2024

Turbulence has associated chaotic features. In the past couple of decades there has been growing interest in the study of these features as an alternative means of understanding turbulent systems. Our own input to this effort has been in contributing to the initial studies of chaos in Eulerian flow using direct numerical simulation (DNS). In this review we discuss the progress achieved in the turbulence community in understanding chaotic measures including our own work. A central relation between turbulence and chaos is one by Ruelle that connects the maximum Lyapunov exponent and the Reynolds number. The first DNS studies, ours amongst them, in obtaining this relation has shown the viability of chaotic simulation studies of Eulerian flow. Such chaotic measures and associated simulation methodology provides an alternative means to probe turbulent flow. Building on this, we have analyzed the finite time Lyapunov exponent (FTLE) and studied its fluctuations, and found that chaotic measures could be quantified accurately even at small simulation box sizes where for comparative sizes spectral measures would be inconclusive. We further highlight applications of chaotic measures in analyzing phase transition behavior in turbulent flow and two dimensional thin layer turbulent systems. This work has shown chaotic measures are an excellent tool that can be used alongside spectral measures in studying turbulent flow., Comment: In press Atmosphere 2024, 18 pages, 7 figures

The LnSbTe (Ln = Lanthanides) family, like isostructural ZrSiS type compounds, has emerged as a fertile playground for exploring the interaction of electronic correlations and magnetic ordering with the nodal line band topology. Here, we report a detailed electronic band structure investigation of TbSbTe, corroborated by electrical transport, thermodynamic, and magnetic studies. Temperature-dependent magnetic susceptibility and thermodynamic transport studies indicate the onset of antiferromagnetic ordering below TN = 5.1 K. The electronic band structure study, carried out with high-resolution angle-resolved photoemission spectroscopy (ARPES) measurements aided with density functional theory based first-principles calculations reveals presence of nodal lines in the GammaX high symmetry direction, forming a diamond-shaped nodal plane around Gamma high symmetry point. A strongly photon energy dependent nodal feature located at the X point of the surface Brillouin zone, indicating an extended nodal line along X R direction, is also observed. This study elucidates the intricate interplay among symmetry-protected band characteristics, the influence of spin orbit coupling, magnetism, and topological properties., Comment: 21 pages, 4 figures

We present Diffusion Soup, a compartmentalization method for Text-to-Image Generation that averages the weights of diffusion models trained on sharded data. By construction, our approach enables training-free continual learning and unlearning with no additional memory or inference costs, since models corresponding to data shards can be added or removed by re-averaging. We show that Diffusion Soup samples from a point in weight space that approximates the geometric mean of the distributions of constituent datasets, which offers anti-memorization guarantees and enables zero-shot style mixing. Empirically, Diffusion Soup outperforms a paragon model trained on the union of all data shards and achieves a 30% improvement in Image Reward (.34 $\to$ .44) on domain sharded data, and a 59% improvement in IR (.37 $\to$ .59) on aesthetic data. In both cases, souping also prevails in TIFA score (respectively, 85.5 $\to$ 86.5 and 85.6 $\to$ 86.8). We demonstrate robust unlearning -- removing any individual domain shard only lowers performance by 1% in IR (.45 $\to$ .44) -- and validate our theoretical insights on anti-memorization using real data. Finally, we showcase Diffusion Soup's ability to blend the distinct styles of models finetuned on different shards, resulting in the zero-shot generation of hybrid styles.

When measuring the Baryon Acoustic Oscillations (BAO) scale from galaxy surveys, one typically assumes a fiducial cosmology when converting redshift measurements into comoving distances and also when defining input parameters for the reconstruction algorithm. A parameterised template for the model to be fitted is also created based on a (possibly different) fiducial cosmology. This model reliance can be considered a form of data compression, and the data is then analysed allowing that the true answer is different from the fiducial cosmology assumed. In this study, we evaluate the impact of the fiducial cosmology assumed in the BAO analysis of the Dark Energy Spectroscopic Instrument (DESI) survey Data Release 1 (DR1) on the final measurements in DESI 2024 III. We utilise a suite of mock galaxy catalogues with survey realism that mirrors the DESI DR1 tracers: the bright galaxy sample (BGS), the luminous red galaxies (LRG), the emission line galaxies (ELG) and the quasars (QSO), spanning a redshift range from 0.1 to 2.1. We compare the four secondary AbacusSummit cosmologies against DESI's fiducial cosmology (Planck 2018). The secondary cosmologies explored include a lower cold dark matter density, a thawing dark energy universe, a higher number of effective species, and a lower amplitude of matter clustering. The mocks are processed through the BAO pipeline by consistently iterating the grid, template, and reconstruction reference cosmologies. We determine a conservative systematic contribution to the error of $0.1\%$ for both the isotropic and anisotropic dilation parameters $\alpha_{\rm iso}$ and $\alpha_{\rm AP}$. We then directly test the impact of the fiducial cosmology on DESI DR1 data., Comment: Supporting publication of DESI 2024 III: Baryon Acoustic Oscillations from Galaxies and Quasars

Large language models have demonstrated remarkable few-shot performance on many natural language understanding tasks. Despite several demonstrations of using large language models in complex, strategic scenarios, there lacks a comprehensive framework for evaluating agents' performance across various types of reasoning found in games. To address this gap, we introduce GameBench, a cross-domain benchmark for evaluating strategic reasoning abilities of LLM agents. We focus on 9 different game environments, where each covers at least one axis of key reasoning skill identified in strategy games, and select games for which strategy explanations are unlikely to form a significant portion of models' pretraining corpuses. Our evaluations use GPT-3 and GPT-4 in their base form along with two scaffolding frameworks designed to enhance strategic reasoning ability: Chain-of-Thought (CoT) prompting and Reasoning Via Planning (RAP). Our results show that none of the tested models match human performance, and at worst GPT-4 performs worse than random action. CoT and RAP both improve scores but not comparable to human levels.

The temporal dynamics of quantitative metrics or key performance indicators (KPIs) are central to decision making within enterprise organizations. Recently, major business intelligence providers have introduced new infrastructure for defining, sharing, and monitoring metric values. However, these values are often presented in isolation and appropriate context is seldom externalized. In this design study, we present RemixTape, an application for constructing structured narratives around metrics. With design imperatives grounded in an formative interview study, RemixTape provides a hierarchical canvas for collecting and coordinating sequences of line chart representations of metrics, along with the ability to externalize situational context around them. RemixTape incorporates affordances to semantically align and annotate juxtaposed charts and text, as well as recommendations of complementary charts based on metrics already present on the canvas. We evaluated RemixTape in a user study in which six enterprise data professionals reproduced and extended partial narratives, with participants appreciating RemixTape as a novel alternative to dashboards, galleries, and slide presentations for supporting conversations about metrics. We conclude with a reflection on how aspects of RemixTape could generalize beyond metrics, with a call to define a conceptual foundation for remixing in the context of visualization., Comment: SUPPLEMENTAL MATERIAL. The scenario video (Sec. 4.4): https://youtu.be/9h6aHvqg9vA; the tutorial video (Sec. 5.2): https://youtu.be/uZPZ5-oiqZk

Fine-tuning large language models is a popular choice among users trying to adapt them for specific applications. However, fine-tuning these models is a demanding task because the user has to examine several factors, such as resource budget, runtime, model size and context length among others. A specific challenge is that fine-tuning is memory intensive, imposing constraints on the required hardware memory and context length of training data that can be handled. In this work, we share a detailed study on a variety of fine-tuning optimizations across different fine-tuning scenarios. In particular, we assess Gradient Checkpointing, Low-Rank Adaptation, DeepSpeed's Zero Redundancy Optimizer and FlashAttention. With a focus on memory and runtime, we examine the impact of different optimization combinations on GPU memory usage and execution runtime during fine-tuning phase. We provide our recommendation on the best default optimization for balancing memory and runtime across diverse model sizes. We share effective strategies for fine-tuning very large models with tens or hundreds of billions of parameters and enabling large context lengths during fine-tuning. Furthermore, we propose the appropriate optimization mixtures for fine-tuning under GPU resource limitations., Comment: 10 pages, 4 figures. Revised text for clarity, updated references

Many problems in computer networking rely on parsing collections of network traces (e.g., traffic prioritization, intrusion detection). Unfortunately, the availability and utility of these collections is limited due to privacy concerns, data staleness, and low representativeness. While methods for generating data to augment collections exist, they often fall short in replicating the quality of real-world traffic In this paper, we i) survey the evolution of traffic simulators/generators and ii) propose the use of state-space models, specifically Mamba, for packet-level, synthetic network trace generation by modeling it as an unsupervised sequence generation problem. Early evaluation shows that state-space models can generate synthetic network traffic with higher statistical similarity to real traffic than the state-of-the-art. Our approach thus has the potential to reliably generate realistic, informative synthetic network traces for downstream tasks., Comment: 7 pages, 3 figures, 4 tables

Deriving meaningful representations from complex, high-dimensional data in unsupervised settings is crucial across diverse machine learning applications. This paper introduces a framework for multi-scale graph network embedding based on spectral graph wavelets that employs a contrastive learning approach. A significant feature of the proposed embedding is its capacity to establish a correspondence between the embedding space and the input feature space which aids in deriving feature importance of the original features. We theoretically justify our approach and demonstrate that, in Paley-Wiener spaces on combinatorial graphs, the spectral graph wavelets operator offers greater flexibility and better control over smoothness properties compared to the Laplacian operator. We validate the effectiveness of our proposed graph embedding on a variety of public datasets through a range of downstream tasks, including clustering and unsupervised feature importance.

We measure the clustering of Lyman Alpha Emitting galaxies (LAEs) selected from the One-hundred-square-degree DECam Imaging in Narrowbands (ODIN) survey, with spectroscopic follow-up from Dark Energy Spectroscopic Instrument (DESI). We use DESI spectroscopy to optimize our selection and to constrain the interloper fraction and redshift distribution of our narrow-band selected sources. We select samples of 4000 LAEs at z=2.45 and 3.1 in 9 sq.deg. centered on the COSMOS field with median LyA fluxes of 10^{-16}erg/s/cm2. Covariances and cosmological inferences are obtained from a series of mock catalogs built upon high-resolution N-body simulations that match the footprint, number density, redshift distribution and observed clustering of the sample. We find that both samples have a correlation length of r_0=(3.0\pm 0.2)Mpc/h. Within our fiducial cosmology these correspond to 3D number densities of 10^{-3} h^3/Mpc^3 and, from our mock catalogs, biases of 1.7 and 2.0 at z=2.45 and 3.1, respectively. We discuss the implications of these measurements for the use of LAEs as large-scale structure tracers for high-redshift cosmology., Comment: 26 pages, 13 figures; minor revisions to match version accepted by JCAP

This study evaluates the effectiveness of child-robot interactions with the HaKsh-E social robot in India, examining both individual and group interaction settings. The research centers on game-based interactions designed to teach hand hygiene to children aged 7-11. Utilizing video analysis, rubric assessments, and post-study questionnaires, the study gathered data from 36 participants. Findings indicate that children in both settings developed positive perceptions of the robot in terms of the robot's trustworthiness, closeness, and social support. The significant difference in the interaction level scores presented in the study suggests that group settings foster higher levels of interaction, potentially due to peer influence and collaborative dynamics. While both settings showed significant improvements in learning outcomes, the individual setting had more pronounced learning gains. This suggests that personal interactions with the robot might lead to deeper or more effective learning experiences. Consequently, this study concludes that individual interaction settings are more conducive for focused learning gains, while group settings enhance interaction and engagement., Comment: 6 pages, 6 figures, Accepted for presentation at ICRAS 2024 (https://www.icras.org/)

Let $X$ be a smooth complex projective variety equipped with an action of a linear algebraic group $G$ over $\mathbb{C}$. Let $D$ be a reduced effective divisor on $X$ that is invariant under the $G$--action on $X$. Let $s_D$ be the canonical section of $\mathcal{O}_X(D)$ vanishing along $D$. Given a positive integer $r$, consider the stack $\mathfrak{X} := \mathfrak{X}_{(\mathcal{O}_X(D),\, s_D,\, r)}$ of $r$-th roots of $(\mathcal{O}_X, s_D)$ together with the natural morphism $\pi : \mathfrak{X} \to X$. Under the assumption that $G$ has no non-trivial characters, we show that the $G$--action on $X$ naturally lifts to a $G$--action on $\mathfrak{X}$ such that $\pi$ become $G$--equivariant, and the tautological invertible sheaf $\mathscr{M}$ on $\mathfrak{X}$ admits a linearization of this $G$--action. Finally, we define the notions of $G$--equivariant logarithmic connections on $\mathfrak{X}$ and $G$--equivariant parabolic connections on $X$ with rational parabolic weights along $D$, and establish an equivalence between the category of $G$--equivariant logarithmic connections on $\mathfrak{X}$ and the category of $G$--equivariant parabolic connections on $X$ with rational parabolic weights along $D$.

Open-weight LLMs are particularly appealing choices to generate training data for fine-tuning Code LLMs on domain-specific service robot applications because they are cost-effective, customizable, and offer better privacy protection. However, unlike proprietary LLMs, open-weight models are more error-prone and often produce programs that violate domain-specific constraints. A promising solution is to incorporate a robot simulator with a well-defined environment to verify program correctness. Yet, these environments require pre-enumeration of relevant entities and their states, which limits the diversity of programs that can be effectively verified. In this work, we introduce ROBO-INSTRUCT that preserves the diversity of programs generated by an LLM while providing the correctness of simulator-based checking. ROBO-INSTRUCT introduces ROBOSIM to dynamically synthesize consistent simulation environments for each generated program. Moreover, ROBO-INSTRUCT handles subtler instruction-program inconsistencies that do not result in a constraint violation via INSTALIGN, an LLM-aided instruction-program alignment process. Given domain-specific APIs and a few seed examples, ROBO-INSTRUCT can leverage an 8B Llama3 model to generate a training dataset for fine-tuning a 7B CodeLlama model. Our fine-tuned model achieves a 28.75% improvement in pass@1 over the original base model and a 13.75% improvement compared to its SELF-INSTRUCT-finetuned counterparts, even surpassing the performance of a few proprietary LLMs, such as GPT-3.5-Turbo and Gemini-Pro.

Personal names simultaneously differentiate individuals and categorize them in ways that are important in a given society. While the natural language processing community has thus associated personal names with sociodemographic characteristics in a variety of tasks, researchers have engaged to varying degrees with the established methodological problems in doing so. To guide future work that uses names and sociodemographic characteristics, we provide an overview of relevant research: first, we present an interdisciplinary background on names and naming. We then survey the issues inherent to associating names with sociodemographic attributes, covering problems of validity (e.g., systematic error, construct validity), as well as ethical concerns (e.g., harms, differential impact, cultural insensitivity). Finally, we provide guiding questions along with normative recommendations to avoid validity and ethical pitfalls when dealing with names and sociodemographic characteristics in natural language processing., Comment: Gender Bias in Natural Language Processing Workshop at ACL 2024

Dark Energy Spectroscopic Instrument (DESI) uses more than 2.4 million Emission Line Galaxies (ELGs) for 3D large-scale structure (LSS) analyses in its Data Release 1 (DR1). Such large statistics enable thorough research on systematic uncertainties. In this study, we focus on spectroscopic systematics of ELGs. The redshift success rate ($f_{\rm goodz}$) is the relative fraction of secure redshifts among all measurements. It depends on observing conditions, thus introduces non-cosmological variations to the LSS. We, therefore, develop the redshift failure weight ($w_{\rm zfail}$) and a per-fibre correction ($\eta_{\rm zfail}$) to mitigate these dependences. They have minor influences on the galaxy clustering. For ELGs with a secure redshift, there are two subtypes of systematics: 1) catastrophics (large) that only occur in a few samples; 2) redshift uncertainty (small) that exists for all samples. The catastrophics represent 0.26\% of the total DR1 ELGs, composed of the confusion between O\,\textsc{ii} and sky residuals, double objects, total catastrophics and others. We simulate the realistic 0.26\% catastrophics of DR1 ELGs, the hypothetical 1\% catastrophics, and the truncation of the contaminated $1.31

We present the technical details on how large-scale structure (LSS) catalogs are constructed from redshifts measured from spectra observed by the Dark Energy Spectroscopic Instrument (DESI). The LSS catalogs provide the information needed to determine the relative number density of DESI tracers as a function of redshift and celestial coordinates and, e.g., determine clustering statistics. We produce catalogs that are weighted subsamples of the observed data, each matched to a weighted `random' catalog that forms an unclustered sampling of the probability density that DESI could have observed those data at each location. Precise knowledge of the DESI observing history and associated hardware performance allows for a determination of the DESI footprint and the number of times DESI has covered it at sub-arcsecond level precision. This enables the completeness of any DESI sample to be modeled at this same resolution. The pipeline developed to create LSS catalogs has been designed to easily allow robustness tests and enable future improvements. We describe how it allows ongoing work improving the match between galaxy and random catalogs, such as including further information when assigning redshifts to randoms, accounting for fluctuations in target density, accounting for variation in the redshift success rate, and accommodating blinding schemes., Comment: Accepted (by JCAP) version of supporting publication of DESI 2024II: Sample definitions, characteristics, and two-point clustering statistics

We use the forward modeling pipeline, Obiwan, to study the imaging systematics of the Luminous Red Galaxies (LRGs) targeted by the Dark Energy Spectroscopic Instrument (DESI). We update the Obiwan pipeline, which had previously been developed to simulate the optical images used to target DESI data, to further simulate WISE images in the infrared. This addition makes it possible to simulate the DESI LRGs sample, which utilizes WISE data in the target selection. Deep DESI imaging data combined with a method to account for biases in their shapes is used to define a truth sample of potential LRG targets. We simulate a total of 15 million galaxies to obtain a simulated LRG sample (Obiwan LRGs) that predicts the variations in target density due to imaging properties. We find that the simulations predict the trends with depth observed in the data, including how they depend on the intrinsic brightness of the galaxies. We observe that faint LRGs are the main contributing power of the imaging systematics trend induced by depth. We also find significant trends in the data against Galactic extinction that are not predicted by Obiwan. These trends depend strongly on the particular map of Galactic extinction chosen to test against, implying Large-Scale Structure systematic contamination (e.g. Cosmic-Infrared Background) in the Galactic extinction maps is a likely root cause. We additionally observe that the DESI LRGs sample exhibits a complex dependency on a combination of seeing, depth, and intrinsic galaxy brightness, which is not replicated by Obiwan, suggesting discrepancies between the current simulation settings and the actual observations. The detailed findings we present should be used to guide any observational systematics mitigation treatment for the clustering of the DESI LRG sample., Comment: 46 pages, 26 figures

The Transiting Exoplanet Survey Satellite (TESS) has discovered hundreds of new worlds, with TESS planet candidates now outnumbering the total number of confirmed planets from $\textit{Kepler}$. Owing to differences in survey design, TESS continues to provide planets that are better suited for subsequent follow-up studies, including mass measurement through radial velocity (RV) observations, compared to Kepler targets. In this work, we present the TESS-Keck Survey's (TKS) Mass Catalog: a uniform analysis of all TKS RV survey data which has resulted in mass constraints for 126 planets and candidate signals. This includes 58 mass measurements that have reached $\geq5\sigma$ precision. We confirm or validate 32 new planets from the TESS mission either by significant mass measurement (15) or statistical validation (17), and we find no evidence of likely false positives among our entire sample. This work also serves as a data release for all previously unpublished TKS survey data, including 9,204 RV measurements and associated activity indicators over our three year survey. We took the opportunity to assess the performance of our survey, and found that we achieved many of our goals including measuring the mass of 38 small ($<4R_{\oplus}$) planets, nearly achieving the TESS mission's basic science requirement. In addition, we evaluated the performance of the Automated Planet Finder (APF) as survey support and observed meaningful constraints on system parameters due to its more uniform phase coverage. Finally, we compared our measured masses to those predicted by commonly used mass-radius relations and investigated evidence of systematic bias., Comment: 51 pages (22 of text), 24 figures

The image matching field has been witnessing a continuous emergence of novel learnable feature matching techniques, with ever-improving performance on conventional benchmarks. However, our investigation shows that despite these gains, their potential for real-world applications is restricted by their limited generalization capabilities to novel image domains. In this paper, we introduce OmniGlue, the first learnable image matcher that is designed with generalization as a core principle. OmniGlue leverages broad knowledge from a vision foundation model to guide the feature matching process, boosting generalization to domains not seen at training time. Additionally, we propose a novel keypoint position-guided attention mechanism which disentangles spatial and appearance information, leading to enhanced matching descriptors. We perform comprehensive experiments on a suite of $7$ datasets with varied image domains, including scene-level, object-centric and aerial images. OmniGlue's novel components lead to relative gains on unseen domains of $20.9\%$ with respect to a directly comparable reference model, while also outperforming the recent LightGlue method by $9.5\%$ relatively.Code and model can be found at https://hwjiang1510.github.io/OmniGlue, Comment: CVPR 2024

Advancements in wireless ultrasound technology allow for point of care cartilage imaging, yet validation against traditional ultrasound units remains to be established for knee cartilage outcomes. Therefore, the purpose of our study was to establish the agreement of articular cartilage thickness and echo-intensity measures between traditional and wireless ultrasound units utilizing automatic-gain and normalization techniques. We used traditional and wireless ultrasound to assess the femoral cartilage via transverse suprapatellar scans with the knee in maximum flexion in 71 female NCAA Division I athletes (age: 20.0$\pm$1.3 years, height: 171.7$\pm$8.7 cm, mass: 69.4$\pm$11.0 kg). Wireless ultrasound images (auto-gain and standard gain) were compared to traditional ultrasound images (standard gain) before and after normalization. Ultrasound image pixel values were algebraically scaled to normalize imaging parameter differences between units. Mean thickness and echo-intensity of the global and sub-regions of interest were measured for unnormalized and normalized images. Intraclass correlation coefficients ($ICC_{2,k}$) for absolute agreement, standard error of the measurement, and minimum detectable difference were calculated between the traditional and wireless ultrasound units across both gain parameters and normalization. Cartilage thickness demonstrated good to excellent agreement for all regions ($ICC_{2,k} = 0.83 {\text -} 0.95$) regardless of gain and normalization. However, mean echo-intensity demonstrated poor to moderate agreement in all regions regardless of gain and normalization ($ICC_{2,k} = 0.23 {\text -} 0.68 $). While there was a high level of agreement between units when assessing cartilage thickness, further research in ultrasound beam forming may lead to improvements in agreement of cartilage echo-intensity between ultrasound units., Comment: 19 Pages, 1 figure, 3 tables

The work introduces a bio-inspired leader-follower system based on an innovative mechanism proposed as software latching that aims to improve collaboration and coordination between a leader agent and the associated autonomous followers. The system utilizes software latching to establish real-time communication and synchronization between the leader and followers. A layered architecture is proposed, encompassing perception, decision-making, and control modules. Challenges such as uncertainty, dynamic environments, and communication latency are addressed using Deep learning and real-time data processing pipelines. The follower robot is equipped with sensors and communication modules that enable it to track and trace the agent of interest or avoid obstacles. The followers track the leader and dynamically avoid obstacles while maintaining a safe distance from it. The experimental results demonstrate the proposed system's effectiveness, making it a promising solution for achieving success in tasks that demand multi-robot systems capable of navigating complex dynamic environments.

A wide range of symbolic analysis and optimization problems can be formalized using polyhedra. Sub-classes of polyhedra, also known as sub-polyhedral domains, are sought for their lower space and time complexity. We introduce the Strided Difference Bound Matrix (SDBM) domain, which represents a sweet spot in the context of optimizing compilers. Its expressiveness and efficient algorithms are particularly well suited to the construction of machine learning compilers. We present decision algorithms, abstract domain operators and computational complexity proofs for SDBM. We also conduct an empirical study with the MLIR compiler framework to validate the domain's practical applicability. We characterize a sub-class of SDBMs that frequently occurs in practice, and demonstrate even faster algorithms on this sub-class., Comment: Preprint and extended from the CAV 2024 conference version. Fixed issue in arxiv version where URLs were not wrapped

Broad absorption line quasars (BALs) exhibit blueshifted absorption relative to a number of their prominent broad emission features. These absorption features can contribute to quasar redshift errors and add absorption to the Lyman-alpha (LyA) forest that is unrelated to large-scale structure. We present a detailed analysis of the impact of BALs on the Baryon Acoustic Oscillation (BAO) results with the LyA forest from the first year of data from the Dark Energy Spectroscopic Instrument (DESI). The baseline strategy for the first year analysis is to mask all pixels associated with all BAL absorption features that fall within the wavelength region used to measure the forest. We explore a range of alternate masking strategies and demonstrate that these changes have minimal impact on the BAO measurements with both DESI data and synthetic data. This includes when we mask the BAL features associated with emission lines outside of the forest region to minimize their contribution to redshift errors. We identify differences in the properties of BALs in the synthetic datasets relative to the observational data, as well as use the synthetic observations to characterize the completeness of the BAL identification algorithm, and demonstrate that incompleteness and differences in the BALs between real and synthetic data also do not impact the BAO results for the LyA forest., Comment: 32 pages, 8 figures, accepted by JCAP

We explore the influence of radio-mode feedback on the properties of the cool circumgalactic medium (CGM). To this end, we assemble a statistical sample of approximately 30,000 radio galaxies with background quasars by combining optical spectroscopic measurements of luminous red galaxies (LRGs) and quasars from the year 1 dataset of Dark Energy Spectroscopic Instrument (DESI) and radio sources from the LOw-Frequency ARray Two-metre Sky Survey (LoTSS) DR2 catalog and the Very Large Array Sky Survey (VLASS) quick look catalog. Galaxies with similar optical properties but with no radio counterparts in LoTSS and VLASS are selected as the control group. We measure the cool CGM properties of radio galaxies and their control samples traced by MgII absorption lines, including covering fraction, rest equivalent width, and gas kinematics. Our results show no significant difference in the properties of gas around radio galaxies and their control sample, indicating that the operating radio-mode feedback of massive galaxies does not produce detectable effects on the properties of the cool CGM. Finally, we show that the CGM of radio galaxies contain a non-negligible amount of cool gas with approximately 10^10 solar masses. This abundance can place a stringent constraint on the radio-mode feedback models., Comment: 20 pages, 12 figures

The popularity of Large Language Models (LLMs) have unleashed a new age ofLanguage Agents for solving a diverse range of tasks. While contemporary frontier LLMs are capable enough to power reasonably good Language agents, the closed-API model makes it hard to improve in cases they perform sub-optimally. To address this, recent works have explored ways to improve their performance using techniques like self-reflection and prompt optimization. Unfortunately, techniques like self-reflection can be used only in an online setup, while contemporary prompt optimization techniques are designed and tested to work on simple tasks. To this end, we introduce MetaReflection, a novel offline reinforcement learning technique that enhances the performance of Language Agents by augmenting a semantic memory based on experiential learnings from past trials. We demonstrate the efficacy of MetaReflection by evaluating across multiple domains, including complex logical reasoning, biomedical semantic similarity, open world question answering, and vulnerability threat detection, in Infrastructure-as-Code, spanning different agent designs. MetaReflection boosts Language agents' performance by 4% to 16.82% over the raw GPT-4 baseline and performs on par with existing state-of-the-art prompt optimization techniques while requiring fewer LLM calls., Comment: We release our experimental code at: https://aka.ms/metareflection-code

Monolithic large language models (LLMs) like GPT-4 have paved the way for modern generative AI applications. Training, serving, and maintaining monolithic LLMs at scale, however, remains prohibitively expensive and challenging. The disproportionate increase in compute-to-memory ratio of modern AI accelerators have created a memory wall, necessitating new methods to deploy AI. Composition of Experts (CoE) is an alternative modular approach that lowers the cost and complexity of training and serving. However, this approach presents two key challenges when using conventional hardware: (1) without fused operations, smaller models have lower operational intensity, which makes high utilization more challenging to achieve; and (2) hosting a large number of models can be either prohibitively expensive or slow when dynamically switching between them. In this paper, we describe how combining CoE, streaming dataflow, and a three-tier memory system scales the AI memory wall. We describe Samba-CoE, a CoE system with 150 experts and a trillion total parameters. We deploy Samba-CoE on the SambaNova SN40L Reconfigurable Dataflow Unit (RDU) - a commercial dataflow accelerator architecture that has been co-designed for enterprise inference and training applications. The chip introduces a new three-tier memory system with on-chip distributed SRAM, on-package HBM, and off-package DDR DRAM. A dedicated inter-RDU network enables scaling up and out over multiple sockets. We demonstrate speedups ranging from 2x to 13x on various benchmarks running on eight RDU sockets compared with an unfused baseline. We show that for CoE inference deployments, the 8-socket RDU Node reduces machine footprint by up to 19x, speeds up model switching time by 15x to 31x, and achieves an overall speedup of 3.7x over a DGX H100 and 6.6x over a DGX A100.

Moderating user-generated content on online platforms is crucial for balancing user safety and freedom of speech. Particularly in the United States, platforms are not subject to legal constraints prescribing permissible content. Each platform has thus developed bespoke content moderation policies, but there is little work towards a comparative understanding of these policies across platforms and topics. This paper presents the first systematic study of these policies from the 43 largest online platforms hosting user-generated content, focusing on policies around copyright infringement, harmful speech, and misleading content. We build a custom web-scraper to obtain policy text and develop a unified annotation scheme to analyze the text for the presence of critical components. We find significant structural and compositional variation in policies across topics and platforms, with some variation attributable to disparate legal groundings. We lay the groundwork for future studies of ever-evolving content moderation policies and their impact on users.

We study the axion portal between the visible and the dark sector, where the Dark Matter is charged under a simple abelian extension of the Standard Model. In general, such models are anomalous and are rendered gauge invariant by a St\"ukelberg axion through Wess-Zumino/Green-Schwarz mechanism. This axion mixes with other Goldstone bosons in the model to give a physical axi-Higgs which becomes massive upon breaking the anomalous gauge group. Such axi-Higgs fields charged under the anomalous symmetry act as mediators for the Dark Matter annihilation to Standard Model particles and can lead to an efficient freeze-out mechanism. Here, we show that the St{\"u}kelberg axion and the resultant axi-Higgs, with its appropriate shift symmetry cancels the quantum anomalies and also generate the observed relic density for the Dark Matter. Moreover, we show that the relevant parameter space in our model, where photon production dominates, is safe from FermiLAT, Cherenkov Telescope Array, and H.E.S.S. indirect detection experiments., Comment: 17 pages, 8 figures

We present the MOST Hosts survey (Multi-Object Spectroscopy of Transient Hosts). The survey is planned to run throughout the five years of operation of the Dark Energy Spectroscopic Instrument (DESI) and will generate a spectroscopic catalog of the hosts of most transients observed to date, in particular all the supernovae observed by most public, untargeted, wide-field, optical surveys (PTF/iPTF, SDSS II, ZTF, DECAT, DESIRT). Scientific questions for which the MOST Hosts survey will be useful include Type Ia supernova cosmology, fundamental plane and peculiar velocity measurements, and the understanding of the correlations between transients and their host galaxy properties. Here, we present the first release of the MOST Hosts survey: 21,931 hosts of 20,235 transients. These numbers represent 36% of the final MOST Hosts sample, consisting of 60,212 potential host galaxies of 38,603 transients (a transient can be assigned multiple potential hosts). Of these galaxies, 40% do not appear in the DESI primary target list and therefore require a specific program like MOST Hosts. Of all the transients in the MOST Hosts list, only 26.7% have existing classifications, and so the survey will provide redshifts (and luminosities) for nearly 30,000 transients. A preliminary Hubble diagram and a transient luminosity-duration diagram are shown as examples of future potential uses of the MOST Hosts survey. The survey will also provide a training sample of spectroscopically observed transients for photometry-only classifiers, as we enter an era when most newly observed transients will lack spectroscopic classification. The MOST Hosts DESI survey data will be released through the Wiserep platform on a rolling cadence and updated to match the DESI releases. Dates of future releases and updates are available through the https://mosthosts.desi.lbl.gov website., Comment: Submitted to ApJS

With the rapid surge in the prevalence of Large Language Models (LLMs), individuals are increasingly turning to conversational AI for initial insights across various domains, including health-related inquiries such as disease diagnosis. Many users seek potential causes on platforms like ChatGPT or Bard before consulting a medical professional for their ailment. These platforms offer valuable benefits by streamlining the diagnosis process, alleviating the significant workload of healthcare practitioners, and saving users both time and money by avoiding unnecessary doctor visits. However, Despite the convenience of such platforms, sharing personal medical data online poses risks, including the presence of malicious platforms or potential eavesdropping by attackers. To address privacy concerns, we propose a novel framework combining FHE and Deep Learning for a secure and private diagnosis system. Operating on a question-and-answer-based model akin to an interaction with a medical practitioner, this end-to-end secure system employs Fully Homomorphic Encryption (FHE) to handle encrypted input data. Given FHE's computational constraints, we adapt deep neural networks and activation functions to the encryted domain. Further, we also propose a faster algorithm to compute summation of ciphertext elements. Through rigorous experiments, we demonstrate the efficacy of our approach. The proposed framework achieves strict security and privacy with minimal loss in performance.