Search

India has made rapid progress towards the universalization of school education, hand in hand with a decline in child labour. Despite progress, child labour persists in the country, just as school attendance and completion rates reveal gaps in educational attainment. This report addresses a timely need for new research to help decision-makers and practitioners build an improved understanding of how to?strengthen the role of education in the elimination of child labour across India. Drawing on quantitative secondary data analysis and qualitative primary data analysis, it offers an accessible and rigorous overview of recent patterns in child labour in India and their linkages with children's school participation. The report also explores children's perspectives of their labour and schooling experiences, including in migration contexts as well as during the COVID-19 pandemic and resulting school closures. The report concludes with programmatic and research recommendations for further action.

Machine learning classifiers often produce probabilistic predictions that are critical for accurate and interpretable decision-making in various domains. The quality of these predictions is generally evaluated with proper losses like cross-entropy, which decompose into two components: calibration error assesses general under/overconfidence, while refinement error measures the ability to distinguish different classes. In this paper, we provide theoretical and empirical evidence that these two errors are not minimized simultaneously during training. Selecting the best training epoch based on validation loss thus leads to a compromise point that is suboptimal for both calibration error and, most importantly, refinement error. To address this, we introduce a new metric for early stopping and hyperparameter tuning that makes it possible to minimize refinement error during training. The calibration error is minimized after training, using standard techniques. Our method integrates seamlessly with any architecture and consistently improves performance across diverse classification tasks.

We show that learning-rate schedules for large model training behave surprisingly similar to a performance bound from non-smooth convex optimization theory. We provide a bound for the constant schedule with linear cooldown; in particular, the practical benefit of cooldown is reflected in the bound due to the absence of logarithmic terms. Further, we show that this surprisingly close match between optimization theory and practice can be exploited for learning-rate tuning: we achieve noticeable improvements for training 124M and 210M Llama-type models by (i) extending the schedule for continued training with optimal learning-rate, and (ii) transferring the optimal learning-rate across schedules.

Power transformers are subjected to electrical currents and temperature fluctuations that, if not properly controlled, can lead to major deterioration of their insulation system. Therefore, monitoring the temperature of a power transformer is fundamental to ensure a long-term operational life. Models presented in the IEC 60076-7 and IEEE standards, for example, monitor the temperature by calculating the top-oil and the hot-spot temperatures. However, these models are not very accurate and rely on the power transformers' properties. This paper focuses on finding an alternative method to predict the top-oil temperatures given previous measurements. Given the large quantities of data available, machine learning methods for time series forecasting are analyzed and compared to the real measurements and the corresponding prediction of the IEC standard. The methods tested are Artificial Neural Networks (ANNs), Time-series Dense Encoder (TiDE), and Temporal Convolutional Networks (TCN) using different combinations of historical measurements. Each of these methods outperformed the IEC 60076-7 model and they are extended to estimate the temperature rise over ambient. To enhance prediction reliability, we explore the application of quantile regression to construct prediction intervals for the expected top-oil temperature ranges. The best-performing model successfully estimates conditional quantiles that provide sufficient coverage.

Inverse Constraint Learning (ICL) is the problem of inferring constraints from safe (i.e., constraint-satisfying) demonstrations. The hope is that these inferred constraints can then be used downstream to search for safe policies for new tasks and, potentially, under different dynamics. Our paper explores the question of what mathematical entity ICL recovers. Somewhat surprisingly, we show that both in theory and in practice, ICL recovers the set of states where failure is inevitable, rather than the set of states where failure has already happened. In the language of safe control, this means we recover a backwards reachable tube (BRT) rather than a failure set. In contrast to the failure set, the BRT depends on the dynamics of the data collection system. We discuss the implications of the dynamics-conditionedness of the recovered constraint on both the sample-efficiency of policy search and the transferability of learned constraints., Comment: 12 pages, 3 figures

I am a person and so are you. Philosophically we sometimes grant personhood to non-human animals, and entities such as sovereign states or corporations can legally be considered persons. But when, if ever, should we ascribe personhood to AI systems? In this paper, we outline necessary conditions for AI personhood, focusing on agency, theory-of-mind, and self-awareness. We discuss evidence from the machine learning literature regarding the extent to which contemporary AI systems, such as language models, satisfy these conditions, finding the evidence surprisingly inconclusive. If AI systems can be considered persons, then typical framings of AI alignment may be incomplete. Whereas agency has been discussed at length in the literature, other aspects of personhood have been relatively neglected. AI agents are often assumed to pursue fixed goals, but AI persons may be self-aware enough to reflect on their aims, values, and positions in the world and thereby induce their goals to change. We highlight open research directions to advance the understanding of AI personhood and its relevance to alignment. Finally, we reflect on the ethical considerations surrounding the treatment of AI systems. If AI systems are persons, then seeking control and alignment may be ethically untenable., Comment: AAAI-25 AI Alignment Track

Accurate rainfall data are crucial for effective climate services, especially in Sub-Saharan Africa, where agriculture heavily depends on rain-fed systems. However, the sparse distribution of rain-gauge networks in the region necessitates reliance on satellite and reanalysis rainfall products (REs) for rainfall estimation. This study evaluated the performance of eight REs -- CHIRPS, TAMSAT, CHIRP, ENACTS, ERA5, AgERA5, PERSIANN-CDR, and PERSIANN-CCS-CDR -- in Zambia and Ghana using a point-to-pixel validation approach. The analysis encompassed spatial consistency, annual rainfall summaries, seasonal patterns, and rainfall intensity detection across 38 ground stations. Results indicated that no single product performed optimally across all contexts, underscoring the need for application-specific recommendations. All products exhibited a high probability of detection (POD) for dry days in Zambia and northern Ghana (with 70% $<$ POD $<$ 100%, and 60% $<$ POD $<$ 85% respectively), suggesting their potential utility for drought-related studies in these areas. Conversely, all products showed limited skill in detecting heavy and violent rains (with POD close to 0%), rendering them unsuitable for extreme rainfall analysis (such as floods) in the current form. Products integrated with station data (ENACTS, CHIRPS, and TAMSAT) outperformed their counterparts under many contexts, highlighting the importance of calibration with local observations. Bias correction is strongly recommended, as varying levels of biases were evident across different rainfall summaries. A critical area for advancement is extreme rainfall detection. Future research should focus on this aspect., Comment: 33 pages, 24 figures

Continuous Integration (CI) consists of an automated build process involving continuous compilation, testing, and packaging of the software system. While CI comes up with several advantages related to quality and time to delivery, CI also presents several challenges addressed by a large body of research. To better understand the literature so as to help practitioners find solutions for their problems and guide future research, we conduct a systematic review of 97 studies on build optimization published between 2006 and 2024, which we summarized according to their goals, methodologies, used datasets, and leveraged metrics. The identified build optimization studies focus on two main challenges: (1) long build durations, and (2) build failures. To meet the first challenge, existing studies have developed a range of techniques, including predicting build outcome and duration, selective build execution, and build acceleration using caching or repairing performance smells. The causes of build failures have been the subject of several studies, leading to the development of techniques for predicting build script maintenance and automating repair. Recent studies have also focused on predicting flaky build failures caused by environmental issues. The majority of these techniques use machine learning algorithms and leverage build metrics, which we classify into five categories. Additionally, we identify eight publicly available build datasets for build optimization research., Comment: An earlier version of this work was submitted to ACM CSUR in November 2023

We present new color-$\Upsilon_*$ (mass-to-light) models to convert WISE W1 fluxes into stellar masses. We outline a range of possible star formation histories and chemical evolution scenarios to explore the confidence limits of stellar population models on the value of $\Upsilon_*$. We conclude that the greatest uncertainties (around 0.1 dex in $\Upsilon_*$) occur for the bluest galaxies with the strongest variation in recent star formation. For high mass galaxies, the greatest uncertainty arises from the proper treatment of bulge/disk separation in which to apply different $\Upsilon_*$ relations appropriate for those differing underlying stellar populations. We compare our deduced stellar masses with those deduced from {\it Spitzer} 3.6$\mu$m fluxes and stellar mass estimates in the literature using optical photometry and different $\Upsilon_*$ modeling. We find the correspondence to be excellent, arguing that rest-frame near-IR photometry is still more advantageous than other wavelengths., Comment: 20 pages, 7 figures, 2 tables, accepted by AJ

Quantum networking and computing technologies demand scalable hardware with high-speed control for large systems of quantum devices. Solid-state platforms have emerged as promising candidates, offering scalable fabrication for a wide range of qubits. Architectures based on spin-photon interfaces allow for highly-connected quantum networks over photonic links, enabling entanglement distribution for quantum networking and distributed quantum computing protocols. With the potential to address these demands, optically-active spin defects in silicon are one proposed platform for building quantum technologies. Here, we electrically excite the silicon T centre in integrated optoelectronic devices that combine nanophotonic waveguides and cavities with p-i-n diodes. We observe single-photon electroluminescence from a cavity-coupled T centre with $g^{(2)}(0)=0.05(2)$. Further, we use the electrically-triggered emission to herald the electron spin state, initializing it with $92(8)\%$ fidelity. This shows, for the first time, electrically-injected single-photon emission from a silicon colour centre and a new method of electrically-triggered spin initialization. These findings present a new telecommunications band light source for silicon and a highly parallel control method for T centre quantum processors, advancing the T centre as a versatile defect for scalable quantum technologies., Comment: 12 pages and 3 figures in main manuscript; 8 pages and 8 figures in supplementary information

The VC-dimension of a family of sets is a measure of its combinatorial complexity used in machine learning theory, computational geometry, and even model theory. Computing the VC-dimension of the $k$-fold union of geometric set systems has been an open and difficult combinatorial problem, dating back to Blumer, Ehrenfeucht, Haussler, and Warmuth in 1989, who ask about the VC-dimension of $k$-fold unions of half-spaces in $\mathbb{R}^d$. Let $\mathcal{F}_1$ denote the family of all lines in $\mathbb{R}^2$. It is well-known that $\mathsf{VC}\text{-}\mathsf{dim}(\mathcal{F}_1) = 2$. In this paper, we study the $2$-fold and $3$-fold unions of $\mathcal{F}_1$, denoted $\mathcal{F}_2$ and $\mathcal{F}_3$, respectively. We show that $\mathsf{VC}\text{-}\mathsf{dim}(\mathcal{F}_2) = 5$ and $\mathsf{VC}\text{-}\mathsf{dim}(\mathcal{F}_3) = 9$. Moreover, we give complete characterisations of the subsets of $\mathbb{R}^2$ of maximal size that can be shattered by $\mathcal{F}_2$ and $\mathcal{F}_3$, showing they are exactly two and five, respectively, up to isomorphism in the language of the point-line incidence relation., Comment: 31 pages, including two appendices

Recent high-throughput experiments unveil substantial electrophysiological diversity among uncoupled healthy myocytes under identical conditions. To quantify inter-cell variability, the values of a subset of the parameters in a well-regarded mathematical model of the action potential of rabbit ventricular myocytes are estimated from fluorescence voltage measurements of a large number of cells. Statistical inference yields a population of nearly 1200 cell-specific model variants that, on a population-level replicate experimentally measured biomarker ranges and distributions, and in contrast to earlier studies, also match experimental biomarker values on a cell-by-cell basis. This model population may be regarded as a random sample from the phenotype of healthy rabbit ventricular myocytes. Uni-variate and bi-variate joint marginal distributions of the estimated parameters are presented, and the parameter dependencies of several commonly utilised electrophysiological biomarkers are revealed. Parameter values are weakly correlated, while summary metrics such as the action potential duration are not strongly dependent on any single electrophysiological characteristic of the myocyte. Our results demonstrate the feasibility of accurately and efficiently fitting entire action potential waveforms at scale. Keywords: cellular excitability, rabbit ventricular myocytes, fluorescence voltage measurements, action potential waveform, parameter estimation in differential equations, noisy time series, Comment: Accepted for publication in Royal Society Open Science (ISSN:2054-5703) on 2025-01-13. Supplementary material available as open source from Royal Society Open Science

Various examples of folding curves in $R^{2}$ have been considered: dragons and other square curves, terdragons and other triangular curves, Peano-Gosper curves based on hexagons. They are self-avoiding. They form coverings of $R^{2}$, by one curve or by a small number of curves, which satisfy the local isomorphism property. They were used to define some fractals. We construct an example with similar properties in $R^{3}$., Comment: 23 pages, 8 figures

Scanning electron microscopy (SEM), a century-old technique, is today a ubiquitous method of imaging the surface of nanostructures. However, most SEM detectors simply count the number of secondary electrons from a material of interest, and thereby overlook the rich material information contained within them. Here, by simple modifications to a standard SEM tool, we resolve the momentum and energy information of secondary electrons by directly imaging the electron plume generated by the electron beam of the SEM. Leveraging these spectroscopic imaging capabilities, our technique is able to image lateral electric fields across a prototypical silicon p-n junctions and to distinguish differently doped regions, even when buried beyond depths typically accessible by SEM. Intriguingly, the sub-surface sensitivity of this technique reveals unexpectedly strong surface band bending within nominally passivated semiconductor structures, providing useful insights for complex layered component designs, in which interfacial dynamics dictate device operation. These capabilities for non-invasive, multi-modal probing of complicated electronic components are crucial in today's electronic manufacturing but is largely inaccessible even with sophisticated techniques. These results show that seemingly simple SEM can be extended to probe complex and useful material properties., Comment: Main text: 16 pages, 5 figure. Supporting information: 18 pages, 14 figures

This research presents an Encoded Spatial Multi-Tier Federated Learning approach for a comprehensive evaluation of aggregated models for geospatial data. In the client tier, encoding spatial information is introduced to better predict the target outcome. The research aims to assess the performance of these models across diverse datasets and spatial attributes, highlighting variations in predictive accuracy. Using evaluation metrics such as accuracy, our research reveals insights into the complexities of spatial granularity and the challenges of capturing underlying patterns in the data. We extended the scope of federated learning (FL) by having multi-tier along with the functionality of encoding spatial attributes. Our N-tier FL approach used encoded spatial data to aggregate in different tiers. We obtained multiple models that predicted the different granularities of spatial data. Our findings underscore the need for further research to improve predictive accuracy and model generalization, with potential avenues including incorporating additional features, refining model architectures, and exploring alternative modeling approaches. Our experiments have several tiers representing different levels of spatial aspects. We obtained accuracy of 75.62% and 89.52% for the global model without having to train the model using the data constituted with the designated tier. The research also highlights the importance of the proposed approach in real-time applications., Comment: IEEE ICCE 2025

Learning the structure of Bayesian networks from data provides insights into underlying processes and the causal relationships that generate the data, but its usefulness depends on the homogeneity of the data population, a condition often violated in real-world applications. In such cases, using a single network structure for inference can be misleading, as it may not capture sub-population differences. To address this, we propose a novel approach of modelling a mixture of Bayesian networks where component probabilities depend on individual characteristics. Our method identifies both network structures and demographic predictors of sub-population membership, aiding personalised interventions. We evaluate our method through simulations and a youth mental health case study, demonstrating its potential to improve tailored interventions in health, education, and social policy.

The continuous delivery of modern software requires the execution of many automated pipeline jobs. These jobs ensure the frequent release of new software versions while detecting code problems at an early stage. For TELUS, our industrial partner in the telecommunications field, reliable job execution is crucial to minimize wasted time and streamline Continuous Deployment (CD). In this context, flaky job failures are one of the main issues hindering CD. Prior studies proposed techniques based on machine learning to automate the detection of flaky jobs. While valuable, these solutions are insufficient to address the waste associated with the diagnosis of flaky failures, which remain largely unexplored due to the wide range of underlying causes. This study examines 4,511 flaky job failures at TELUS to identify the different categories of flaky failures that we prioritize based on Recency, Frequency, and Monetary (RFM) measures. We identified 46 flaky failure categories that we analyzed using clustering and RFM measures to determine 14 priority categories for future automated diagnosis and repair research. Our findings also provide valuable insights into the evolution and impact of these categories. The identification and prioritization of flaky failure categories using RFM analysis introduce a novel approach that can be used in other contexts., Comment: This paper has been accepted at the 47th International Conference on Software Engineering: Software Engineering in Practice 2025

''Positive geometries'' are a class of semi-algebraic domains which admit a unique ''canonical form'': a logarithmic form whose residues match the boundary structure of the domain. The study of such geometries is motivated by recent progress in particle physics, where the corresponding canonical forms are interpreted as the integrands of scattering amplitudes. We recast these concepts in the language of mixed Hodge theory, and identify ''genus zero pairs'' of complex algebraic varieties as a natural and general framework for the study of positive geometries and their canonical forms. In this framework, we prove some basic properties of canonical forms which have previously been proved or conjectured in the literature. We give many examples and study in detail the case of arrangements of hyperplanes and convex polytopes., Comment: 51 pages, 11 figures

This paper introduces indefinite proximities inherent in the collection of physical objects found in a dynamical system. Axiomatically, these indefinite proximities lead to a new form of Hausdorff topology, which is indefinite descriptively. The main results in this paper are (1) Every descriptive proximity space on a dynamical system is indefinite (Theorem 1), (2) Every dynamical system has an indefinite descriptive Hausdorff topology (Theorem 3), and (3) The energy of a dynamical system varies with every clock tick (Theorem 4). An application of these results is given in terms of the detection of those portions of a dynamical system that are stable and that have low energy dissipation., Comment: 15 pages, 6 figures

The total deuterium abundance [D/H] in the universe is set by just two processes: the creation of deuterium in Big Bang Nucleosynthesis at an abundance of [D/H]$=2.58\pm0.13\times10^{-5}$, and its destruction within stellar interiors. Measurements of the total [D/H] abundance can potentially provide a probe of Galactic chemical evolution, however, most measurements of [D/H] are only sensitive to the gas-phase deuterium, and the amount of deuterium sequestered in carbonaceous dust grains is debated. With the launch of JWST, it is now possible to measure the gas-phase [D/H] at unprecedented sensitivity and distances through observation of mid-IR lines of H$_2$ and HD. We employ data from the JWST Observations of Young protoStars (JOYS) program to measure the gas-phase [D/H] abundance with a rotation diagram analysis towards 5 nearby low-mass and 5 distant high-mass protostellar outflows. The gas-phase [D/H] varies between low-mass sources by up to a factor of $\sim4$, despite these sources likely having formed in a region of the Galactic disk that would be expected to have nearly constant total [D/H]. Most measurements of gas-phase [D/H] from our work or previous studies produce [D/H] $\lesssim 1.0\times10^{-5}$, a factor of $2-4$ lower than found from local UV absorption lines and as expected from Galactic chemical evolution models. The variations in [D/H] between our low-mass sources and the low [D/H] with respect to Galactic chemical evolution models suggest that our observations are not sensitive to the total [D/H]. Significant depletion of deuterium onto carbonaceous dust grains is a possible explanation, and tentative evidence of enhanced [D/H] towards shock positions with higher gas-phase Fe abundance is seen in the HH 211 outflow. Deeper observations of HD and H$_2$ in shocked environments and modelling of dust-grain destruction are warranted to test for the effects of depletion., Comment: 30 Pages, 20 Figures. Accepted for publication in Astronomy and Astrophysics

Neuromorphic computing, inspired by the brain's parallel and energy-efficient processing, offers a transformative approach to artificial intelligence. In this study, we fabricated optimized spin-transfer torque nano-oscillators (STNOs) and investigated their dynamic behaviors using a hybrid excitation scheme combining AC laser illumination and DC bias currents. Laser-induced thermal gradients generate pulsed thermoelectric voltages ($V_{\text{AC}}$) via the Tunnel Magneto-Seebeck (TMS) effect, while the addition of bias currents enhances this response, producing both $V_{\text{AC}}$ and a DC component ($V_{\text{DC}}$). Magnetic field sweeps reveal distinct switching between parallel (P) and antiparallel (AP) magnetization states in both voltage components, supporting multistate memory applications. Millivolt-range thermovoltage signals in open-circuit conditions demonstrate CMOS compatibility, enabling simplified, scalable neuromorphic systems. Under biased conditions, enhanced thermovoltage outputs exhibit intriguing phenomena, including spikes correlated with Barkhausen jumps and double-switching behavior, offering insights into magnetization dynamics and vortex transitions. These features resemble neural spiking behavior, suggesting applications in spiking neural networks, reservoir computing, multistate logic, analog computing, and high-resolution sensing. By bridging spintronic phenomena with practical applications, this work provides a versatile platform for next-generation AI technologies and adaptive computing architectures.

Given the prospects of the low-altitude economy (LAE) and the popularity of unmanned aerial vehicles (UAVs), there are increasing demands on monitoring flying objects at low altitude in wide urban areas. In this work, the widely deployed long-term evolution (LTE) base station (BS) is exploited to illuminate UAVs in bistatic trajectory tracking. Specifically, a passive sensing receiver with two digital antenna arrays is proposed and developed to capture both the line-of-sight (LoS) signal and the scattered signal off a target UAV. From their cross ambiguity function, the bistatic range, Doppler shift and angle-of-arrival (AoA) of the target UAV can be detected in a sequence of time slots. In order to address missed detections and false alarms of passive sensing, a multi-target tracking framework is adopted to track the trajectory of the target UAV. It is demonstrated by experiments that the proposed UAV tracking system can achieve a meter-level accuracy., Comment: 13 pages, 10 figures, submitted to IEEE Journal for possible publication

Contaminant observations and outliers often cause problems when estimating the parameters of cognitive models, which are statistical models representing cognitive processes. In this study, we test and improve the robustness of parameter estimation using amortized Bayesian inference (ABI) with neural networks. To this end, we conduct systematic analyses on a toy example and analyze both synthetic and real data using a popular cognitive model, the Drift Diffusion Models (DDM). First, we study the sensitivity of ABI to contaminants with tools from robust statistics: the empirical influence function and the breakdown point. Next, we propose a data augmentation or noise injection approach that incorporates a contamination distribution into the data-generating process during training. We examine several candidate distributions and evaluate their performance and cost in terms of accuracy and efficiency loss relative to a standard estimator. Introducing contaminants from a Cauchy distribution during training considerably increases the robustness of the neural density estimator as measured by bounded influence functions and a much higher breakdown point. Overall, the proposed method is straightforward and practical to implement and has a broad applicability in fields where outlier detection or removal is challenging.

Speech brain-computer interfaces aim to decipher what a person is trying to say from neural activity alone, restoring communication to people with paralysis who have lost the ability to speak intelligibly. The Brain-to-Text Benchmark '24 and associated competition was created to foster the advancement of decoding algorithms that convert neural activity to text. Here, we summarize the lessons learned from the competition ending on June 1, 2024 (the top 4 entrants also presented their experiences in a recorded webinar). The largest improvements in accuracy were achieved using an ensembling approach, where the output of multiple independent decoders was merged using a fine-tuned large language model (an approach used by all 3 top entrants). Performance gains were also found by improving how the baseline recurrent neural network (RNN) model was trained, including by optimizing learning rate scheduling and by using a diphone training objective. Improving upon the model architecture itself proved more difficult, however, with attempts to use deep state space models or transformers not yet appearing to offer a benefit over the RNN baseline. The benchmark will remain open indefinitely to support further work towards increasing the accuracy of brain-to-text algorithms.

Hanbury Brown and Twiss (HBT) interferometry is a milestone experiment that transformed our understanding of the nature of light. The concept was demonstrated in 1956 to measure the radii of stars through photon coincidence detection. This form of coincidence detection later became a cornerstone of modern quantum optics. Here we connect HBT interferometry to the physics of scintillation, the process of spontaneous light emission upon excitation by high-energy particles, such as x-rays. Our work reveals intrinsic photon bunching in the scintillation process, which we utilize to elucidate its underlying light emission mechanisms. Specifically, g^((2) ) ({\tau}) enables the quantitative extraction of scintillation lifetime and light yield, showing their dependence on temperature and X-ray flux as well. This approach provides a characterization method that we benchmark on a wide gamut of scintillators, including rare-earth-doped garnets and perovskite nanocrystals. Our method is particularly important for nano- and micro-scale scintillators, whose properties are challenging to quantify by conventional means: We extract the scintillation properties in perovskite nanocrystals of only a few hundreds of nanometers, observing strong photon bunching (g^((2) ) (0)>50). Our research paves the way for broader use of photon-coincidence measurement and methods from quantum optics in studying materials with complex optical properties in extremes regions of the electromagnetic spectrum.

The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought before answering has the potential to unlock substantial benefits, while also increasing potential risks that stem from heightened intelligence. Our results underscore the need for building robust alignment methods, extensively stress-testing their efficacy, and maintaining meticulous risk management protocols. This report outlines the safety work carried out for the OpenAI o1 and OpenAI o1-mini models, including safety evaluations, external red teaming, and Preparedness Framework evaluations.

Non-reversible lifts reduce the relaxation time of reversible diffusions at most by a square root. For reversible diffusions on domains in Euclidean space, or, more generally, on a Riemannian manifold with boundary, non-reversible lifts are in particular given by the Hamiltonian flow on the tangent bundle, interspersed with random velocity refreshments, or perturbed by Ornstein-Uhlenbeck noise, and reflected at the boundary. In order to prove that for certain choices of parameters, these lifts achieve the optimal square-root reduction up to a constant factor, precise upper bounds on relaxation times are required. A key tool for deriving such bounds by space-time Poincar\'e inequalities is a quantitative space-time divergence lemma. Extending previous work of Cao, Lu and Wang, we establish such a divergence lemma with explicit constants for general locally convex domains with smooth boundary in Riemannian manifolds satisfying a lower, not necessarily positive, curvature bound. As a consequence, we prove optimality of the lifts described above up to a constant factor, provided the deterministic transport part of the dynamics and the noise are adequately balanced. Our results show for example that an integrated Ornstein-Uhlenbeck process on a locally convex domain with diameter $d$ achieves a relaxation time of the order $d$, whereas, in general, the Poincar\'e constant of the domain is of the order $d^2$., Comment: 41 pages

An optimal transport (OT) problem seeks to find the cheapest mapping between two distributions with equal total density, given the cost of transporting density from one place to another. Unbalanced OT allows for different total density in each distribution. This is the typical setting for precipitation forecast and observation data, when considering the densities as accumulated rainfall, or intensity. In this work, entropic unbalanced OT and its associated Sinkhorn divergence are examined as a spatial forecast verification method for precipitation data. It offers many attractive features, such as morphing one field into another, defining a distance between fields and providing feature based optimal assignment. It is found that the Sinkhorn divergence is robust against the common double penalty problem (a form of phase error), aligns with expert assessments of model performance, and allows for a variety of novel pictorial illustrations of error. It provides informative summary scores, and has few limitations to its application. Combined, these findings place unbalanced entropy regularised optimal transport and the Sinkhorn divergence as an informative method which follows geometric intuition., Comment: submitted to Meteorological Applications

Sliced inverse regression (SIR) is a popular sufficient dimension reduction method that identifies a few linear transformations of the covariates without losing regression information with the response. In high-dimensional settings, SIR can be combined with sparsity penalties to achieve sufficient dimension reduction and variable selection simultaneously. Nevertheless, both classical and sparse estimators assume the covariates are exogenous. However, endogeneity can arise in a variety of situations, such as when variables are omitted or are measured with error. In this article, we show such endogeneity invalidates SIR estimators, leading to inconsistent estimation of the true central subspace. To address this challenge, we propose a two-stage Lasso SIR estimator, which first constructs a sparse high-dimensional instrumental variables model to obtain fitted values of the covariates spanned by the instruments, and then applies SIR augmented with a Lasso penalty on these fitted values. We establish theoretical bounds for the estimation and selection consistency of the true central subspace for the proposed estimators, allowing the number of covariates and instruments to grow exponentially with the sample size. Simulation studies and applications to two real-world datasets in nutrition and genetics illustrate the superior empirical performance of the two-stage Lasso SIR estimator compared with existing methods that disregard endogeneity and/or nonlinearity in the outcome model., Comment: 63 pages

Autoregressive models, built based on the Next Token Prediction (NTP) paradigm, show great potential in developing a unified framework that integrates both language and vision tasks. In this work, we rethink the NTP for autoregressive image generation and propose a novel Next Patch Prediction (NPP) paradigm. Our key idea is to group and aggregate image tokens into patch tokens containing high information density. With patch tokens as a shorter input sequence, the autoregressive model is trained to predict the next patch, thereby significantly reducing the computational cost. We further propose a multi-scale coarse-to-fine patch grouping strategy that exploits the natural hierarchical property of image data. Experiments on a diverse range of models (100M-1.4B parameters) demonstrate that the next patch prediction paradigm could reduce the training cost to around 0.6 times while improving image generation quality by up to 1.0 FID score on the ImageNet benchmark. We highlight that our method retains the original autoregressive model architecture without introducing additional trainable parameters or specifically designing a custom image tokenizer, thus ensuring flexibility and seamless adaptation to various autoregressive models for visual generation., Comment: Code: https://github.com/PKU-YuanGroup/Next-Patch-Prediction, v2: add related work "Patch-Level Training for Large Language Models"

Robot learning is witnessing a significant increase in the size, diversity, and complexity of pre-collected datasets, mirroring trends in domains such as natural language processing and computer vision. Many robot learning methods treat such datasets as multi-task expert data and learn a multi-task, generalist policy by training broadly across them. Notably, while these generalist policies can improve the average performance across many tasks, the performance of generalist policies on any one task is often suboptimal due to negative transfer between partitions of the data, compared to task-specific specialist policies. In this work, we argue for the paradigm of training policies during deployment given the scenarios they encounter: rather than deploying pre-trained policies to unseen problems in a zero-shot manner, we non-parametrically retrieve and train models directly on relevant data at test time. Furthermore, we show that many robotics tasks share considerable amounts of low-level behaviors and that retrieval at the "sub"-trajectory granularity enables significantly improved data utilization, generalization, and robustness in adapting policies to novel problems. In contrast, existing full-trajectory retrieval methods tend to underutilize the data and miss out on shared cross-task content. This work proposes STRAP, a technique for leveraging pre-trained vision foundation models and dynamic time warping to retrieve sub-sequences of trajectories from large training corpora in a robust fashion. STRAP outperforms both prior retrieval algorithms and multi-task learning methods in simulated and real experiments, showing the ability to scale to much larger offline datasets in the real world as well as the ability to learn robust control policies with just a handful of real-world demonstrations., Comment: Project website at https://weirdlabuw.github.io/strap/

Weak lensing surveys, along with most other late-Universe probes, have consistently measured a lower amplitude of the matter fluctuation spectrum, denoted by the parameter $S_8$, compared to predictions from early-Universe measurements in cosmic microwave background data. Improper modelling of nonlinear scales may partially explain these discrepancies in lensing surveys. This study investigates whether the conventional approach to addressing small-scale biases remains optimal for Stage-IV lensing surveys. We demonstrate that conventional weak lensing estimators are affected by scale leakage from theoretical biases at nonlinear scales, which influence all observed scales. Using the BNT transform, we propose an $\ell$-cut methodology that effectively controls this leakage. The BNT transform reorganizes weak lensing data in $\ell$ space, aligning it with $k$ space, thereby reducing the mixing of nonlinear scales and providing a more accurate interpretation of the data. We evaluate the BNT approach by comparing HMcode, Halofit, Baryon Correction Model and AxionHMcode mass power spectrum models using Euclid-like survey configurations. Additionally, we introduce a new estimator to quantify scale leakage in both the BNT and noBNT approaches. Our findings show that BNT outperforms traditional methods, preserving cosmological constraints while significantly mitigating theoretical biases., Comment: 21 pages, 10 figures, 3 tables, comments welcome!

In Embodied Question Answering (EQA), agents must explore and develop a semantic understanding of an unseen environment in order to answer a situated question with confidence. This remains a challenging problem in robotics, due to the difficulties in obtaining useful semantic representations, updating these representations online, and leveraging prior world knowledge for efficient exploration and planning. Aiming to address these limitations, we propose GraphEQA, a novel approach that utilizes real-time 3D metric-semantic scene graphs (3DSGs) and task relevant images as multi-modal memory for grounding Vision-Language Models (VLMs) to perform EQA tasks in unseen environments. We employ a hierarchical planning approach that exploits the hierarchical nature of 3DSGs for structured planning and semantic-guided exploration. Through experiments in simulation on the HM-EQA dataset and in the real world in home and office environments, we demonstrate that our method outperforms key baselines by completing EQA tasks with higher success rates and fewer planning steps., Comment: Project website: https://saumyasaxena.github.io/grapheqa

Recently, Large Language Models (LLMs) have shown impressive performance in character understanding tasks, such as analyzing the roles, personalities, and relationships of fictional characters. However, the extensive pre-training corpora used by LLMs raise concerns that they may rely on memorizing popular fictional works rather than genuinely understanding and reasoning about them. In this work, we argue that 'gist memory'-capturing essential meaning - should be the primary mechanism for character understanding tasks, as opposed to 'verbatim memory' - exact match of a string. We introduce a simple yet effective method to mitigate mechanized memorization in character understanding evaluations while preserving the essential implicit cues needed for comprehension and reasoning. Our approach reduces memorization-driven performance on popular fictional works from 96% accuracy to 72% and results in up to an 18% drop in accuracy across various character understanding tasks. These findings underscore the issue of data contamination in existing benchmarks, which often measure memorization rather than true character understanding.

The transverse momentum dependent (TMD) soft function and Collins-Soper (CS) kernel may be obtained by formulating the Wilson line in terms of auxiliary one-dimensional fermion fields on the lattice. Our computation takes place in the region of the lattice that corresponds to the "spacelike" region in Minkowski space in order to obtain the Collins soft function. The matching of our result to the Collins soft function is achieved through the mapping of the auxiliary field directional vector that to the Wilson line rapidity. In Euclidean space, this directional vector is complex, having a purely imaginary time component. We present some exploratory numerical results of our lattice calculation, and discuss the methodology employed., Comment: Submitted to Proceedings of Science for The 41st International Symposium on Lattice Field Theory (Lattice 2024)

Time series data plays a critical role across diverse domains such as healthcare, energy, and finance, where tasks like classification, anomaly detection, and forecasting are essential for informed decision-making. Recently, large language models (LLMs) have gained prominence for their ability to handle complex data and extract meaningful insights. This study investigates whether LLMs are effective for time series data analysis by comparing their performance with non-LLM-based approaches across three tasks: classification, anomaly detection, and forecasting. Through a series of experiments using GPT4TS and autoregressive models, we evaluate their performance on benchmark datasets and assess their accuracy, precision, and ability to generalize. Our findings indicate that while LLM-based methods excel in specific tasks like anomaly detection, their benefits are less pronounced in others, such as forecasting, where simpler models sometimes perform comparably or better. This research highlights the role of LLMs in time series analysis and lays the groundwork for future studies to systematically explore their applications and limitations in handling temporal data.

Resource allocation is fundamental for cloud systems to ensure efficient resource sharing among tenants. However, the scale of such optimization problems has outgrown the capabilities of commercial solvers traditionally employed in production. To scale up resource allocation, prior approaches either tailor solutions to specific problems or rely on assumptions tied to particular workloads. In this work, we revisit real-world resource allocation problems and uncover a common underlying structure: a vast majority of these problems are inherently separable, i.e., they optimize the aggregate utility of individual resource and demand allocations, under separate constraints for each resource and each demand. Building on this insight, we develop DeDe, a general, scalable, and theoretically grounded framework for accelerating resource allocation through a "decouple and decompose" approach. DeDe systematically decouples entangled resource and demand constraints, thereby decomposing the overall optimization into alternating per-resource and per-demand allocations, which can then be solved efficiently and in parallel. We have implemented DeDe as a library extension to an open-source solver, maintaining a familiar user interface. Experimental results across three prominent resource allocation tasks -- traffic engineering, cluster scheduling, and load balancing -- demonstrate DeDe's substantial speedups and robust allocation quality.

We unify the variational hypocoercivity framework established by D. Albritton, S. Armstrong, J.-C. Mourrat, and M. Novack, with the notion of second-order lifts of reversible diffusion processes, recently introduced by A. Eberle and F. L\"orler. We give an abstract, yet fully constructive, presentation of the theory, so that it can be applied to a large class of linear kinetic equations. As this hypocoercivity technique does not twist the reference norm, we can recover accurate and sharp convergence rates in various models. Among those, adaptive Langevin dynamics (ALD) is discussed in full detail and we show that for near-quadratic potentials, with suitable choices of parameters, it is a near-optimal second-order lift of the overdamped Langevin dynamics. As a further consequence, we observe that the Generalised Langevin Equation (GLE) is a also a second-order lift, as the standard (kinetic) Langevin dynamics are, of the overdamped Langevin dynamics. Then, convergence of (GLE) cannot exceed ballistic speed, i.e.~the square root of the rate of the overdamped regime. We illustrate this phenomenon with explicit computations in a benchmark Gaussian case., Comment: 17 pages

Middleware, third-party software intermediaries between users and platforms, has been broached as a means to decentralize the power of social media platforms and enhance user agency. Middleware may enable a more user-centric and democratic approach to shaping digital experiences, offering a flexible architecture as an alternative to both centrally controlled, opaque platforms and an unmoderated, uncurated internet. The widespread adoption of open middleware has long hinged on the cooperation of established major platforms; however, the recent growth of federated platforms, such as Mastodon and Bluesky, has led to increased offerings and user awareness. In this report we consider the potential of middleware as a means of enabling greater user control over curation and moderation - two aspects of the social media experience that are often mired in controversy. We evaluate the trade-offs and negative externalities it might create, and discuss the technological, regulatory, and market dynamics that could either support or hinder its implementation., Comment: 51 pages

Surgical workflow anticipation is the task of predicting the timing of relevant surgical events from live video data, which is critical in Robotic-Assisted Surgery (RAS). Accurate predictions require the use of spatial information to model surgical interactions. However, current methods focus solely on surgical instruments, assume static interactions between instruments, and only anticipate surgical events within a fixed time horizon. To address these challenges, we propose an adaptive graph learning framework for surgical workflow anticipation based on a novel spatial representation, featuring three key innovations. First, we introduce a new representation of spatial information based on bounding boxes of surgical instruments and targets, including their detection confidence levels. These are trained on additional annotations we provide for two benchmark datasets. Second, we design an adaptive graph learning method to capture dynamic interactions. Third, we develop a multi-horizon objective that balances learning objectives for different time horizons, allowing for unconstrained predictions. Evaluations on two benchmarks reveal superior performance in short-to-mid-term anticipation, with an error reduction of approximately 3% for surgical phase anticipation and 9% for remaining surgical duration anticipation. These performance improvements demonstrate the effectiveness of our method and highlight its potential for enhancing preparation and coordination within the RAS team. This can improve surgical safety and the efficiency of operating room usage., Comment: Accepted by IEEE Transactions on Medical Robotics and Bionics, the direct link to the IEEE page will be updated upon publication

Human emotions can be conveyed through nuanced touch gestures. However, there is a lack of understanding of how consistently emotions can be conveyed to robots through touch. This study explores the consistency of touch-based emotional expression toward a robot by integrating tactile and auditory sensory reading of affective haptic expressions. We developed a piezoresistive pressure sensor and used a microphone to mimic touch and sound channels, respectively. In a study with 28 participants, each conveyed 10 emotions to a robot using spontaneous touch gestures. Our findings reveal a statistically significant consistency in emotion expression among participants. However, some emotions obtained low intraclass correlation values. Additionally, certain emotions with similar levels of arousal or valence did not exhibit significant differences in the way they were conveyed. We subsequently constructed a multi-modal integrating touch and audio features to decode the 10 emotions. A support vector machine (SVM) model demonstrated the highest accuracy, achieving 40% for 10 classes, with "Attention" being the most accurately conveyed emotion at a balanced accuracy of 87.65%.

Determining the internal structure of Uranus is a key objective for planetary science. Knowledge of Uranus's bulk composition and the distribution of elements is crucial to understanding its origin and evolutionary path. In addition, Uranus represents a poorly understood class of intermediate-mass planets (intermediate in size between the relatively well studied terrestrial and gas giant planets), which appear to be very common in the Galaxy. As a result, a better characterization of Uranus will also help us to better understand exoplanets in this mass and size regime. Recognizing the importance of Uranus, a Keck Institute for Space Studies (KISS) workshop was held in September 2023 to investigate how we can improve our knowledge of Uranus's internal structure in the context of a future Uranus mission that includes an orbiter and a probe. The scientific goals and objectives of the recently released Planetary Science and Astrobiology Decadal Survey were taken as our starting point. We reviewed our current knowledge of Uranus's interior and identified measurement and other mission requirements for a future Uranus spacecraft, providing more detail than was possible in the Decadal Survey's mission study and including new insights into the measurements to be made. We also identified important knowledge gaps to be closed with Earth-based efforts in the near term that will help guide the design of the mission and interpret the data returned., Comment: Study Report prepared for the W. M. Keck Institute for Space Studies (KISS). Study title: Determining the Interior Structure of Uranus: A Case Study for Leveraging Cross-Discipline Science to Answer Tough Questions. Study dates: September 11-15, 2023. Team Leads: Mark Hofstadter, Ravit Helled, and David Stevenson

We present the first numerical investigation of the method proposed in Ref. [1] to utilize gradient flow to obtain precise determinations of higher moments of PDFs from lattice QCD, circumventing power divergent mixing with lower dimensional operators. We apply this method to obtain moments of the isovector PDF of the pion using four Stabilized Wilson Fermion ensembles with $m_{\pi}\simeq 411$ MeV and lattice spacings $a \simeq 0.064, 0.077, 0.094$, and $0.12$ fm. We present preliminary results of ratios of three-point functions as a function of flow time, which can be used to extract the ratios $\left\langle x^2 \right\rangle/\left\langle x \right\rangle$ and $\left\langle x^3 \right\rangle/\left\langle x \right\rangle$. We find that a significantly higher precision can be achieved with this method compared to the canonical approach, which requires boosting and cannot reach higher than the $\left\langle x^3 \right\rangle$ moment., Comment: 15 pages, 2 figures, contributions to the proceedings of the 41st International Symposium on Lattice Field Theory (LATTICE2024)

We demonstrate the advantages of THz frequency combs for high-resolution spectroscopy. This benefits from wide spectral coverage and the exact knowledge of the frequency position of each comb component. Heterodyne detection combined with a fast Fourier spectrometer enables rapid and simultaneous measurement of more than 80 frequency comb modes covering a 7.5 GHz bandwidth. A spectrum is obtained in under 20 minutes yielding a uniform resolution of 70 kHz. This new setup has been validated by recording more than 150 lines of methanol around 723 GHz, and represents a new solution to exploit THz frequency combs for high-resolution spectroscopy.

High resolution rotational Terahertz (THz) spectroscopy has been widely applied to the studies of numerous polar gas phase molecules, in particular volatile organic compounds (VOCs). During the storage of foodstuffs packed under a protective atmosphere, microbial activity will lead to the generation of a complex mixture of trace gases that could be used as food spoilage indicators. Here we have demonstrated that the THz instrumentation presently available provides sufficient sensitivity and selectivity to monitor the generation of hydrogen sulfide (H2S) in the headspace of packed Atlantic salmon (Salmo salar) fillet portions. A comprehensive comparison was made by selective-ion flow-tube mass spectrometry (SIFT-MS) in order to validate the THz measurements and protocol. The detectivity of a range of alternative compounds for this application is also provided, based on the experimental detection limit observed and molecular spectroscopic properties. Molecules like ethanol, methyl mercaptan and ammonia are suitable indicators with the presently available sensitivity levels, while dimethyl sulfide, acetone and butanone may be considered with a sensitivity improvement of 2 orders of magnitude.

Recently, extended short-term precipitation nowcasting struggles with decreasing precision because of insufficient consideration of meteorological knowledge, such as weather fronts which significantly influence precipitation intensity, duration, and spatial distribution. Therefore, in this paper, we present DuoCast, a novel dual-probabilistic meteorology-aware model designed to address both broad weather evolution and micro-scale fluctuations using two diffusion models, PrecipFlow and MicroDynamic, respectively. Our PrecipFlow model captures evolution trends through an Extreme Precipitation-Aware Encoder (EPA-Encoder), which includes AirConvolution and FrontAttention blocks to process two levels of precipitation data: general and extreme. The output conditions a UNet-based diffusion to produce prediction maps enriched with weather front information. The MicroDynamic model further refines the results to capture micro-scale variability. Extensive experiments on four public benchmarks demonstrate the effectiveness of our DuoCast, achieving superior performance over state-of-the-art methods. Our code is available at https://github.com/ph-w2000/DuoCast.