Search

Ground-based high-resolution cross-correlation spectroscopy (HRCCS; R >~ 15,000) is a powerful complement to space-based studies of exoplanet atmospheres. By resolving individual spectral lines, HRCCS can precisely measure chemical abundance ratios, directly constrain atmospheric dynamics, and robustly probe multidimensional physics. But the subtleties of HRCCS datasets -- e.g., the lack of exoplanetary spectra visible by eye and the statistically complex process of telluric removal -- can make interpreting them difficult. In this work, we seek to clarify the uncertainty budget of HRCCS with a forward-modeling approach. We present a HRCCS observation simulator, scope (https://github.com/arjunsavel/scope), that incorporates spectral contributions from the exoplanet, star, tellurics, and instrument. This tool allows us to control the underlying dataset, enabling controlled experimentation with complex HRCCS methods. Simulating a fiducial hot Jupiter dataset (WASP-77Ab emission with IGRINS), we first confirm via multiple tests that the commonly used principal components analysis does not bias the planetary signal when few components are used. Furthermore, we demonstrate that mildly varying tellurics and moderate wavelength solution errors induce only mild decreases in HRCCS detection significance. However, limiting-case, strongly varying tellurics can bias the retrieved velocities and gas abundances. Additionally, in the low-SNR limit, constraints on gas abundances become highly non-Gaussian. Our investigation of the uncertainties and potential biases inherent in HRCCS data analysis enables greater confidence in scientific results from this maturing method., Comment: 21 pages, 11 figures. Submitted to AJ; comments welcome

Neutrino flavor evolution is critical for understanding the physics of dense astrophysical regimes, including core-collapse supernovae (CCSN). Powerful numerical integration codes exist for simulating these environments, yet a complete understanding of the inherent nonlinearity of collective neutrino flavor oscillations and how it fits within the overall framework of these simulations remains an open challenge. For this reason, we continue developing statistical data assimilation (SDA) to infer solutions to the flavor field in a CCSN envelope, given simulated measurements far from the source. SDA is an inference paradigm designed to optimize a model with sparse data. Our model consists of neutrino beams emanating from a CCSN and coherently interacting with each other and with a background of other matter particles in one dimension $r$. One model feature of high interest is the distribution of those matter particles as a function of radius $r$, or the "matter potential" $V(r)$ -- as it significantly dictates flavor evolution. In this paper, we expand the model beyond previous incarnations, by replacing the monotonically-decaying analytic form for $V(r)$ we previously used with a more complex -- and more physically plausible -- set of profiles derived from a one-dimensional (spherically symmetric) hydrodynamics simulation of a CCSN explosion. We ask whether the SDA procedure can use simulated flavor measurements at physically accessible locations (i.e. in vacuum) to determine the extent to which different matter density profiles through which the neutrinos propagate in the matter-dominated regime are compatible with these measurements. Within the scope of our small-scale model, we find that the neutrino flavor measurements in the vacuum regime are able to discriminate between different matter profiles, and we discuss implications regarding a future galactic CCSN detection.

Time-varying inhomogeneities on stellar surfaces constitute one of the largest sources of radial velocity (RV) error for planet detection and characterization. We show that stellar variations, because they manifest on coherent, rotating surfaces, give rise to changes that are complex but useably compact and coherent in the spectral domain. Methods for disentangling stellar signals in RV measurements benefit from modeling the full domain of spectral pixels. We simulate spectra of spotted stars using starry and construct a simple spectrum projection space that is sensitive to the orientation and size of stellar surface features. Regressing measured RVs in this projection space reduces RV scatter by 60-80% while preserving planet shifts. We note that stellar surface variability signals do not manifest in spectral changes that are purely orthogonal to a Doppler shift or exclusively asymmetric in line profiles; enforcing orthogonality or focusing exclusively on asymmetric features will not make use of all the information present in the spectra. We conclude with a discussion of existing and possible implementations on real data based on the presented compact, coherent framework for stellar signal mitigation., Comment: 19 pages, 12 figures, accepted to ApJ

Despite growing interest in incorporating feedback to improve language models, most efforts focus only on sequence-level annotations. In this work, we explore the potential of utilizing fine-grained span-level annotations from offline datasets to improve model quality. We develop a simple finetuning algorithm, called Training with Annotations (TWA), to directly train machine translation models on such annotated data. TWA utilizes targeted span-level error information while also flexibly learning what to penalize within a span. Moreover, TWA considers the overall trajectory of a sequence when deciding which non-error spans to utilize as positive signals. Experiments on English-German and Chinese-English machine translation show that TWA outperforms baselines such as Supervised FineTuning on sequences filtered for quality and Direct Preference Optimization on pairs constructed from the same data.

Purpose: To quantify the relative performance of step counting algorithms in studies that collect free-living high-resolution wrist accelerometry data and to highlight the implications of using these algorithms in translational research. Methods: Five step counting algorithms (four open source and one proprietary) were applied to the publicly available, free-living, high-resolution wrist accelerometry data collected by the National Health and Nutrition Examination Survey (NHANES) in 2011-2014. The mean daily total step counts were compared in terms of correlation, predictive performance, and estimated hazard ratios of mortality. Results: The estimated number of steps were highly correlated (median=0.91, range 0.77 to 0.98), had high and comparable predictive performance of mortality (median concordance=0.72, range 0.70 to 0.73). The distributions of the number of steps in the population varied widely (mean step counts range from 2,453 to 12,169). Hazard ratios of mortality associated with a 500-step increase per day varied among step counting algorithms between HR=0.88 and 0.96, corresponding to a 300% difference in mortality risk reduction ([1-0.88]/[1-0.96]=3). Conclusion: Different step counting algorithms provide correlated step estimates and have similar predictive performance that is better than traditional predictors of mortality. However, they provide widely different distributions of step counts and estimated reductions in mortality risk for a 500-step increase., Comment: 41 pages, 3 figures

We show that a two variable Nevanlinna representation for rational Pick functions can be viewed as the Cauchy transform of a simple undirected graph with variables assigned to the vertices. We give relations between representing functions of certain products of such graphs by way of Schur complements applied to adjacency matrices. We also study the connection between the structure of the graph and the regularity of the representing function at a boundary singularity.

IRAC data have long implied that early ($z\gtrsim6$) galaxies often have very high specific star formation rates (sSFR$\gtrsim$30 Gyr$^{-1}$), but JWST data have shown that at least some early galaxies are forming stars far less vigorously. Here, we systematically analyze the recent star formation histories (SFHs) of a large ($N=368$) sample of $z\sim6$ Lyman-break galaxies (LBGs) spanning $-22\lesssim M_\mathrm{UV}\lesssim-16$ assembled from ACS+NIRCam imaging in the GOODS and Abell 2744 fields. We find that very low H$\alpha$-to-UV luminosity ratios ($L_\mathrm{H\alpha}/L_\mathrm{UV}$) and strong recent downturns in star formation rate (SFR) are $\approx$5$\times$ more common among the UV-faintest subset of our sample ($\langle M_\mathrm{UV}\rangle=-17.4$) compared to the brightest subset ($\langle M_\mathrm{UV}\rangle=-20.0$). The frequency of high $L_\mathrm{H\alpha}/L_\mathrm{UV}$ and strong recent SFR upturns is approximately constant with UV luminosity. We discuss how bursty SFHs naturally reproduce this much greater diversity in recent SFHs among very UV-faint galaxies. Using public NIRSpec/prism data, we newly confirm recent strong SFR downturns among three LBGs in our sample, and validate our photometric inferences on key SFH signatures among $z\sim6$ LBGs in general. Our results imply that early galaxies frequently cycle through phases of rapid stellar mass assembly and other periods of much slower growth. This yields huge ($\gtrsim$1-2 mag) fluctuations in $M_\mathrm{UV}$ on rapid ($\sim$10-30 Myr) timescales, helping explain the surprising abundance of $z>10$ galaxies. Finally, we caution that this burstiness causes all existing high-redshift samples (particularly line-selected samples) to be far less complete to galaxies with long recent phases of low sSFR than those currently undergoing a burst., Comment: Submitted to ApJ. 29 pages, 13 figures, 1 appendix

Lanthanide-doped Y$_{2}$SiO$_{5}$ microcrystals were prepared using the solution combustion, solid state and sol-gel synthesis techniques. Of these, the sol-gel method yields the most reliable and high-quality X2 phase Y$_{2}$SiO$_{5}$ microcrystals. Absorption and laser site-selective fluorescence measurements of Nd$^{3+}$, Eu$^{3+}$ and Er$^{3+}$ doped material, performed at cryogenic temperatures, indicate that the as-grown microcrystals are of high optical quality with inhomogeneously broadened optical linewidths that are comparable to bulk crystals at similar dopant concentrations.

While large language models (LLMs) exhibit significant utility across various domains, they simultaneously are susceptible to exploitation for unethical purposes, including academic misconduct and dissemination of misinformation. Consequently, AI-generated text detection systems have emerged as a countermeasure. However, these detection mechanisms demonstrate vulnerability to evasion techniques and lack robustness against textual manipulations. This paper introduces back-translation as a novel technique for evading detection, underscoring the need to enhance the robustness of current detection systems. The proposed method involves translating AI-generated text through multiple languages before back-translating to English. We present a model that combines these back-translated texts to produce a manipulated version of the original AI-generated text. Our findings demonstrate that the manipulated text retains the original semantics while significantly reducing the true positive rate (TPR) of existing detection methods. We evaluate this technique on nine AI detectors, including six open-source and three proprietary systems, revealing their susceptibility to back-translation manipulation. In response to the identified shortcomings of existing AI text detectors, we present a countermeasure to improve the robustness against this form of manipulation. Our results indicate that the TPR of the proposed method declines by only 1.85% after back-translation manipulation. Furthermore, we build a large dataset of 720k texts using eight different LLMs. Our dataset contains both human-authored and LLM-generated texts in various domains and writing styles to assess the performance of our method and existing detectors. This dataset is publicly shared for the benefit of the research community.

We study the problem of finding fair allocations -- EF1 and EFX -- of indivisible goods with orientations. In an orientation, every agent gets items from their own predetermined set. For EF1, we show that EF1 orientations always exist when agents have monotone valuations, via a pseudopolynomial-time algorithm. This surprisingly positive result is the main contribution of our paper. We complement this result with a comprehensive set of scenarios where our algorithm, or a slight modification of it, finds an EF1 orientation in polynomial time. For EFX, we focus on the recently proposed graph instances, where every agent corresponds to a vertex on a graph and their allowed set of items consists of the edges incident to their vertex. It was shown that finding an EFX orientation is NP-complete in general. We prove that it remains intractable even when the graph has a vertex cover of size 8, or when we have a multigraph with only 10 vertices. We essentially match these strong negative results with a fixed-parameter tractable algorithm that is virtually the best someone could hope for., Comment: 21 pages, 3 figures

Meteorites and their components exhibit a diverse range of oxygen isotope compositions, and the isotopic exchange timescale between dust grains and ambient gas is a key parameter for understanding the spatiotemporal evolution of the solar nebula. As dust grains existed as macroscopic aggregates in the solar nebula, it is necessary to consider the isotopic exchange timescales for these aggregates. Here, we theoretically estimate the isotope exchange timescales between dust aggregates and ambient vapor. The isotope exchange process between aggregates and ambient vapor is divided into four processes: (i) supply of gas molecules to the aggregate surface, (ii) diffusion of molecules within the aggregate, (iii) isotope exchange on the surface of constituent particles, and (iv) isotope diffusion within the particles. We evaluate these timescales and assess which one becomes the rate-determining step. We reveal that the isotope exchange timescale is approximately the same as that of the constituent particles when the aggregate radius is smaller than the critical value, which is a few centimeters when considering the exchange reaction between amorphous forsterite aggregates and water vapor., Comment: 7 pages, 3 figures. Accepted for publication in ApJ

Logistics and transportation networks require a large amount of resources to realize necessary connections between locations and minimizing these resources is a vital aspect of planning research. Since such networks have dynamic connections that are only available at specific times, intricate models are needed to portray them accurately. In this paper, we study the problem of minimizing the number of resources needed to realize a dynamic network, using the temporal graphs model. In a temporal graph, edges appear at specific points in time. Given a temporal graph and a natural number k, we ask whether we can cover every temporal edge exactly once using at most k temporal journeys; in a temporal journey consecutive edges have to adhere to the order of time. We conduct a thorough investigation of the complexity of the problem with respect to four dimensions: (a) whether the type of the temporal journey is a walk, a trail, or a path; (b) whether the chronological order of edges in the journey is strict or non-strict; (c) whether the temporal graph is directed or undirected; (d) whether the start and end points of each journey are given or not. We almost completely resolve the complexity of all these problems and provide dichotomies for each one of them with respect to k.

Radial velocity (RV) follow-up is a critical complement of transiting exoplanet surveys like the Transiting Exoplanet Survey Satellite (TESS ), both for validating discoveries of exoplanets and measuring their masses. Stellar activity introduces challenges to interpreting these measurements because the noise from the host star, which is often correlated in time, can result in high RV uncertainty. A robust understanding of stellar activity and how its timescales interact with the observing cadence can optimize limited RV resources. For this reason, in the era of over-subscribed, high-precision RV measurements, folding stellar activity timescales into the scheduling of observation campaigns is ideal. We present gaspery, an open-source code implementation to enable the optimization of RV observing strategies. Gaspery employs a generalized formulation of the Fisher Information for RV time series that also incorporates information about stellar correlated noise. We show that the information contained in an observing strategy can be significantly affected by beat frequencies between the orbital period of the planet, the stellar rotation period, and the observation epochs. We investigate how the follow-up observing strategy will affect the resulting radial velocity uncertainty, as a function of stellar properties such as the spot decay timescale and rotation period. We then describe two example use cases for gaspery: 1) calculating the minimum number of observations to reach an uncertainty tolerance in a correlated noise regime and 2) finding an optimal strategy given a fixed observing budget. Finally, we outline a prescription for selecting an observing strategy that is generalizable to different targets., Comment: 19 pages, 7 figures, accepted for publication by AAS Journals

Artificial Intelligence (AI) has paved the way for revolutionary decision-making processes, which if harnessed appropriately, can contribute to advancements in various sectors, from healthcare to economics. However, its black box nature presents significant ethical challenges related to bias and transparency. AI applications are hugely impacted by biases, presenting inconsistent and unreliable findings, leading to significant costs and consequences, highlighting and perpetuating inequalities and unequal access to resources. Hence, developing safe, reliable, ethical, and Trustworthy AI systems is essential. Our team of researchers working with Trustworthy and Responsible AI, part of the Transdisciplinary Scholarship Initiative within the University of Calgary, conducts research on Trustworthy and Responsible AI, including fairness, bias mitigation, reproducibility, generalization, interpretability, and authenticity. In this paper, we review and discuss the intricacies of AI biases, definitions, methods of detection and mitigation, and metrics for evaluating bias. We also discuss open challenges with regard to the trustworthiness and widespread application of AI across diverse domains of human-centric decision making, as well as guidelines to foster Responsible and Trustworthy AI models., Comment: 44 pages, 2 figures

In a changing climate, sustainable agriculture is essential for food security and environmental health. However, it is challenging to understand the complex interactions among its biophysical, social, and economic components. Predictive machine learning (ML), with its capacity to learn from data, is leveraged in sustainable agriculture for applications like yield prediction and weather forecasting. Nevertheless, it cannot explain causal mechanisms and remains descriptive rather than prescriptive. To address this gap, we propose causal ML, which merges ML's data processing with causality's ability to reason about change. This facilitates quantifying intervention impacts for evidence-based decision-making and enhances predictive model robustness. We showcase causal ML through eight diverse applications that benefit stakeholders across the agri-food chain, including farmers, policymakers, and researchers.

Quantum computing has tremendous potential to overcome some of the fundamental limitations present in classical information processing. Yet, today's technological limitations in the quality and scaling prevent exploiting its full potential. Quantum computing based on superconducting quantum processing units (QPUs) is among the most promising approaches towards practical quantum advantage. In this article the basic technological approach of IQM Quantum Computers is described covering both the QPU and the rest of the full-stack quantum computer. In particular, the focus is on a 20-qubit quantum computer featuring the Garnet QPU and its architecture, which we will scale up to 150 qubits. We also present QPU and system-level benchmarks, including a median 2-qubit gate fidelity of 99.5% and genuinely entangling all 20 qubits in a Greenberger-Horne-Zeilinger (GHZ) state.

An obstetric goal for a laboring mother is to achieve a vaginal delivery as it reduces the risks inherent in major abdominal surgery (i.e., a Cesarean section). Various medical interventions may be used by a physician to increase the likelihood of this occurring while minimizing maternal and fetal morbidity. However, patients with prelabor rupture of membranes (PROM) have only two commonly used options for cervical ripening, Pitocin and misoprostol. Little research exists on the benefits/risks for these two key drugs for PROM patients. A major limitation with most induction-of-labor related research is the inability to account for differences in \textit{Bishop scores} that are commonly used in obstetrical practice to determine the next induction agent offered to the patient. This creates a confounding factor, which biases the results, but has not been realized in the literature. In this work, we use a Bayesian model of the relationships between the relevant factors, informed by expert physicians, to separate the confounding variable from its actual impact. In doing so, we provide strong evidence that pitocin and buccal misoprostol are equally effective and safe; thus, physicians have more choice in clinical care than previously realized. This is particularly important for developing countries where neither medication may be readily available, and prior guidelines may create an artificial barrier to needed medication., Comment: To appear in the 2024 IEEE 11th International Conference on Data Science and Advanced Analytics (DSAA)

For solar-type stars, spots and their associated magnetic regions induce radial velocity perturbations through the Doppler rotation signal and the suppression of convective blueshift -- collectively known as rotation-modulation. We developed the Rotation-Convection (RC) model: a method of detrending and characterizing rotation-modulation, using only cross-correlation functions or 1-dimensional spectra, without the need for continuous high cadence measurements. The RC method uses a simple model for the anomalous radial velocity induced by an active region and has two inputs: stellar flux (or a flux proxy) and the relative radial velocity between strongly and weakly absorbed wavelengths (analogous to the bisector-inverse slope). On NEID solar data (three month baseline), the RC model lowers the amplitude of rotationally-modulated stellar activity to below the meter-per-second level. For the standard star HD 26965, the RC model detrends the activity signal to the meter-per-second level for HARPS, EXPRES, and NEID observations, even though the temporal density and timespan of the observations differs by an order of magnitude between the three datasets. In addition to detrending, the RC model also characterizes the rotation-modulation signal. From comparison with the Solar Dynamics Observatory, we confirmed that the model accurately recovers and separates the rotation and convection radial velocity components. We also mapped the amplitude of the rotation and convection perturbations as a function of height within the stellar atmosphere. Probing stellar atmospheres with our revised spot model will fuel future innovations in stellar activity mitigation, enabling robust exoplanet detection., Comment: Accepted to AJ, 22 pages, 9 figures

We introduce Imagen 3, a latent diffusion model that generates high quality images from text prompts. We describe our quality and responsibility evaluations. Imagen 3 is preferred over other state-of-the-art (SOTA) models at the time of evaluation. In addition, we discuss issues around safety and representation, as well as methods we used to minimize the potential harm of our models.

The recruitment of teenagers for usable privacy and security research is challenging, but essential. This case study presents our experience using the online flier distribution service Peachjar to recruit minor teenagers for an online security experiment. By distributing fliers to 90 K-12 schools, we recruited a diverse sample of 55 participants at an estimated cost per participant of $43.18. We discuss the benefits and drawbacks of Peachjar, concluding that it can facilitate the recruitment of a geographically diverse sample of teens for online studies, but it requires careful design to protect against spam and may be more expensive than other online methods. We conclude by proposing ways of using Peachjar more effectively., Comment: To be presented at the 9th Workshop on Inclusive Privacy and Security (WIPS 2024) at the Twentieth Symposium on Usable Privacy and Security (SOUPS 2024)

Gallium arsenide optical resonators have been identified as platforms for light-matter interactions ranging from optomechanics to quantum electrodynamics involving nonlinear optics. Here, we present a 5-mm-diameter whispering gallery mode resonator made of undoped gallium arsenide. The fabrication was done using single-point diamond turning followed by polishing with diamond solutions of varying grain size. The resulting resonator was optically characterized to have a quality factor of more than a million.

The escalating impacts of climate change and the increasing demand for sustainable development and natural resource management necessitate innovative technological solutions. Quantum computing (QC) has emerged as a promising tool with the potential to revolutionize these critical areas. This review explores the application of quantum machine learning and optimization techniques for climate change prediction and enhancing sustainable development. Traditional computational methods often fall short in handling the scale and complexity of climate models and natural resource management. Quantum advancements, however, offer significant improvements in computational efficiency and problem-solving capabilities. By synthesizing the latest research and developments, this paper highlights how QC and quantum machine learning can optimize multi-infrastructure systems towards climate neutrality. The paper also evaluates the performance of current quantum algorithms and hardware in practical applications and presents realistic cases, i.e., waste-to-energy in anaerobic digestion, disaster prevention in flooding prediction, and new material development for carbon capture. The integration of these quantum technologies promises to drive significant advancements in achieving climate resilience and sustainable development.

The signal induced by a temperate, terrestrial planet orbiting a Sun-like star is an order of magnitude smaller than the host stars' intrinsic variability. Understanding stellar activity is, therefore, a fundamental obstacle in confirming the smallest exoplanets. We present the Lowell Observatory Solar Telescope (LOST), a solar feed for the EXtreme PREcision Spectrometer (EXPRES) at the 4.3-m Lowell Discovery Telescope (LDT). EXPRES is one of the newest high-resolution spectrographs that accurately measure extreme radial velocity. With LOST/EXPRES, we observe disk-integrated sunlight autonomously throughout the day. In clear conditions, we achieve a ~137,500 optical spectrum of the Sun with a signal-to-noise of 500 in ~150s. Data is reduced using the standard EXPRES pipeline with minimal modification to ensure the data are comparable to the observations of other stars with the LDT. During the first three years of operation, we find a daily RMS of 71 cm/s. Additionally, having two EPRV spectrometers located in Arizona gives us an unprecedented opportunity to benchmark the performance of these planet-finders. We find a RMS of just 55 cm/s when comparing data taken simultaneously with EXPRES and NEID., Comment: SPIE Astronomical Telescopes & Instrumentation proceedings paper

We prove several results about integral versions of Fourier duality for abelian schemes, making use of Pappas's work on integral Grothendieck-Riemann-Roch. If $S$ is smooth quasi-projective of dimension $d$ over a field and $\pi \colon X\to S$ is a $g$-dimensional abelian scheme, we prove, under very mild assumptions on $X/S$, that all classical results about Fourier duality, including the existence of a Beauville decomposition, are valid for the Chow ring $\mathrm{CH}(X;\Lambda)$ with coefficients in the ring $\Lambda = \mathbb{Z}[1/(2g+d+1)!]$. If $X$ admits a polarization $\theta$ of degree $\nu(\theta)^2$ we further construct an $\mathfrak{sl}_2$-action on $\mathrm{CH}(X;\Lambda_\theta)$ with $\Lambda_\theta = \Lambda[1/\nu(\theta)]$, and we show that $\mathrm{CH}(X;\Lambda_\theta)$ is a sum of copies of the symmetric powers $\mathrm{Sym}^n(\mathrm{St})$ of the $2$-dimensional standard representation, for $n=0,\ldots,g$. For an abelian variety over an algebraically closed field, we use our results to produce torsion classes in $\mathrm{CH}^i(X;\Lambda_\theta)$ for every $i\in \{1,\ldots,g\}$., Comment: 22 pages

Carbon dioxide will likely need to be removed from the atmosphere to avoid significant future warming and climate change. Technologies are being developed to remove large quantities of carbon from the atmosphere. Enhanced rock weathering (ERW), where fine-grained silicate minerals are spread on soil, is a promising carbon removal method that can also support crop yields and maintain overall soil health. Quantifying the amount of carbon removed by ERW is crucial for understanding the potential of ERW globally and for building trust in commercial operations. However, reliable and scalable quantification in complex media like soil is challenging and there is not yet a consensus on the best method of doing so. Here we discuss mass-balance methods, where stocks of base cations in soil are monitored over time to infer the amount of inorganic carbon brought into solution by weathering reactions. First, we review the fundamental concepts of mass-balance methods and explain different ways of approaching the mass-balance problem. Then we discuss experimental planning and data collection, suggesting some best practices. Next, we present a software package designed to facilitate a range of tasks in ERW like uncertainty analysis, planning field trials, and validating statistical methods. Finally, we briefly review ways of estimating carbon removal using mass balance before discussing some advantages of Bayesian inference in this context and presenting an example Bayesian model. The model is fit to simulated data and recovers the correct answer with a clear representation of uncertainty.

3D Gaussian Splatting (3DGS) is a promising technique for 3D reconstruction, offering efficient training and rendering speeds, making it suitable for real-time applications.However, current methods require highly controlled environments (no moving people or wind-blown elements, and consistent lighting) to meet the inter-view consistency assumption of 3DGS. This makes reconstruction of real-world captures problematic. We present SpotLessSplats, an approach that leverages pre-trained and general-purpose features coupled with robust optimization to effectively ignore transient distractors. Our method achieves state-of-the-art reconstruction quality both visually and quantitatively, on casual captures. Additional results available at: https://spotlesssplats.github.io

Generative models of language exhibit impressive capabilities but still place non-negligible probability mass over undesirable outputs. In this work, we address the task of updating a model to avoid unwanted outputs while minimally changing model behavior otherwise, a challenge we refer to as a minimal targeted update. We first formalize the notion of a minimal targeted update and propose a method to achieve such updates using negative examples from a model's generations. Our proposed Targeted Negative Training (TNT) results in updates that keep the new distribution close to the original, unlike existing losses for negative signal which push down probability but do not control what the updated distribution will be. In experiments, we demonstrate that TNT yields a better trade-off between reducing unwanted behavior and maintaining model generation behavior than baselines, paving the way towards a modeling paradigm based on iterative training updates that constrain models from generating undesirable outputs while preserving their impressive capabilities., Comment: Published in Transactions of Machine Learning Research

Large Language Models (LLMs) have become extremely potent instruments with exceptional capacities for comprehending and producing human-like text in a wide range of applications. However, the increasing size and complexity of LLMs present significant challenges in both training and deployment, leading to substantial computational and storage costs as well as heightened energy consumption. In this paper, we provide a review of recent advancements and research directions aimed at addressing these challenges and enhancing the efficiency of LLM-based systems. We begin by discussing algorithm-level acceleration techniques focused on optimizing LLM inference speed and resource utilization. We also explore LLM-hardware co-design strategies with a vision to improve system efficiency by tailoring hardware architectures to LLM requirements. Further, we delve into LLM-to-accelerator compilation approaches, which involve customizing hardware accelerators for efficient LLM deployment. Finally, as a case study to leverage LLMs for assisting circuit design, we examine LLM-aided design methodologies for an important task: High-Level Synthesis (HLS) functional verification, by creating a new dataset that contains a large number of buggy and bug-free codes, which can be essential for training LLMs to specialize on HLS verification and debugging. For each aspect mentioned above, we begin with a detailed background study, followed by the presentation of several novel solutions proposed to overcome specific challenges. We then outline future research directions to drive further advancements. Through these efforts, we aim to pave the way for more efficient and scalable deployment of LLMs across a diverse range of applications., Comment: This is an expanded and more comprehensive study based on our invited DAC-24 paper with the same title and co-authors

A fundamental objective in robot manipulation is to enable models to comprehend visual scenes and execute actions. Although existing robot Multimodal Large Language Models (MLLMs) can handle a range of basic tasks, they still face challenges in two areas: 1) inadequate reasoning ability to tackle complex tasks, and 2) high computational costs for MLLM fine-tuning and inference. The recently proposed state space model (SSM) known as Mamba demonstrates promising capabilities in non-trivial sequence modeling with linear inference complexity. Inspired by this, we introduce RoboMamba, an end-to-end robotic MLLM that leverages the Mamba model to deliver both robotic reasoning and action capabilities, while maintaining efficient fine-tuning and inference. Specifically, we first integrate the vision encoder with Mamba, aligning visual data with language embedding through co-training, empowering our model with visual common sense and robot-related reasoning. To further equip RoboMamba with action pose prediction abilities, we explore an efficient fine-tuning strategy with a simple policy head. We find that once RoboMamba possesses sufficient reasoning capability, it can acquire manipulation skills with minimal fine-tuning parameters (0.1\% of the model) and time (20 minutes). In experiments, RoboMamba demonstrates outstanding reasoning capabilities on general and robotic evaluation benchmarks. Meanwhile, our model showcases impressive pose prediction results in both simulation and real-world experiments, achieving inference speeds 7 times faster than existing robot MLLMs. Our project web page: https://sites.google.com/view/robomamba-web

Bioacoustic research, vital for understanding animal behavior, conservation, and ecology, faces a monumental challenge: analyzing vast datasets where animal vocalizations are rare. While deep learning techniques are becoming standard, adapting them to bioacoustics remains difficult. We address this with animal2vec, an interpretable large transformer model, and a self-supervised training scheme tailored for sparse and unbalanced bioacoustic data. It learns from unlabeled audio and then refines its understanding with labeled data. Furthermore, we introduce and publicly release MeerKAT: Meerkat Kalahari Audio Transcripts, a dataset of meerkat (Suricata suricatta) vocalizations with millisecond-resolution annotations, the largest labeled dataset on non-human terrestrial mammals currently available. Our model outperforms existing methods on MeerKAT and the publicly available NIPS4Bplus birdsong dataset. Moreover, animal2vec performs well even with limited labeled data (few-shot learning). animal2vec and MeerKAT provide a new reference point for bioacoustic research, enabling scientists to analyze large amounts of data even with scarce ground truth information., Comment: Code available at: https://github.com/livingingroups/animal2vec | Dataset available at: https://doi.org/10.17617/3.0J0DYB

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

Preference learning algorithms (e.g., RLHF and DPO) are frequently used to steer LLMs to produce generations that are more preferred by humans, but our understanding of their inner workings is still limited. In this work, we study the conventional wisdom that preference learning trains models to assign higher likelihoods to more preferred outputs than less preferred outputs, measured via ranking accuracy. Surprisingly, we find that most state-of-the-art preference-tuned models achieve a ranking accuracy of less than 60% on common preference datasets. We furthermore derive the idealized ranking accuracy that a preference-tuned LLM would achieve if it optimized the DPO or RLHF objective perfectly. We demonstrate that existing models exhibit a significant alignment gap -- i.e., a gap between the observed and idealized ranking accuracies. We attribute this discrepancy to the DPO objective, which is empirically and theoretically ill-suited to fix even mild ranking errors in the reference model, and derive a simple and efficient formula for quantifying the difficulty of learning a given preference datapoint. Finally, we demonstrate that ranking accuracy strongly correlates with the empirically popular win rate metric when the model is close to the reference model used in the objective, shedding further light on the differences between on-policy (e.g., RLHF) and off-policy (e.g., DPO) preference learning algorithms., Comment: NeurIPS 2024 camera-ready