Search

Star-forming clumps have been found to significantly influence the star formation of gas-rich $z>1$ galaxies. Using public data from JWST/NIRCam (COSMOS-Web) and ALMA (FMOS-COSMOS), we study a sample of 32 massive ($>10^{10.5}\,\rm M_{\odot}$) main-sequence galaxies at $z_{\rm spec}\sim1.5$ with $\sim0.3\,\rm kpc$ resolution. We create composite morphological models consisting of bulge, disk, and clumps to fully 'deconstruct' the galaxy images. With the resulting measurements of the flux and size of these components, we find the following: (I)The combined contribution of clumps is $1-30\%$ towards the net star formation rate (SFR) of the host while contributing $1-20\%$ to its stellar mass. The clumps show a correlation between their stellar mass and SFR, but have an increased specific-SFR (sSFR) relative to the star-forming main sequence, with offsets ranging from $0\lesssim\Delta\log\rm sSFR\lesssim 0.4$. They feature star formation surface densities of $10^{-2}-10^{2}\,\rm M_{\odot}/yr/kpc^{2}$, consistent with values observed in local star-forming and starburst galaxies. (II)The clumps span a large range of characteristic sizes ($r_{e}\sim0.1-1\,\rm kpc$) and stellar masses ($\sim 10^{8.0-9.5}\,\rm M_{\odot}$). We estimate a mass-size relation ($r_{e}\propto\rm M_{\star}^{\,0.52\pm0.07}$) along with a stellar mass function (slope, $\alpha=-1.85\pm 0.19$), both suggesting a hierarchical nature similar to that expected in star-forming regions in local galaxies. (III)Our measurements agree with the properties of stellar clumps in $z\gtrsim1$ lensed systems, bridging the gap between lensed and unlensed studies by detecting structures at sub-kpc scales.(IV)Clumps are found to be preferentially located along spiral features visible primarily in the residual rest-frame near-IR images. In conclusion, we present an observation-based, coherent picture of star-forming clumps at $z>1$., Comment: Published in MNRAS; 22 pages, 24 figures

Recent JWST observations have revealed the prevalence of spiral structures at $z > 1$. Unlike in the local Universe, the origin and the consequence of spirals at this epoch remain unexplored. We use public JWST/NIRCam data from the COSMOS-Web survey to map spiral structures in eight massive ($> 10^{10.5}\,\rm M_{\odot}$) star-forming galaxies at $z_{\rm spec} \sim 1.5$. We present a method for systematically quantifying spiral arms at $z>1$, enabling direct measurements of flux distributions. Using rest-frame near-IR images, we construct morphological models accurately tracing spiral arms. We detect offsets ($\sim 0.2 - 0.8\,\rm kpc$) between the rest-frame optical and near-IR flux distributions across most arms. Drawing parallels to the local Universe, we conclude that these offsets reflect the presence of density waves. For nine out of eighteen arms, the offsets indicate spiral shocks triggered by density waves. Five arms have offsets in the opposite direction and are likely associated with tidal interactions. For the remaining cases with no detected offsets, we suggest that stochastic 'clumpy' star formation is the primary driver of their formation. In conclusion, we find a multi-faceted nature of spiral arms at $z > 1$, similar to that in the local Universe., Comment: Accepted for publication in ApJ Letters; 12 pages, 6 figures

We present results from a high cadence multi-wavelength observational campaign of the enigmatic changing look AGN 1ES 1927+654 from May 2022- April 2024, coincident with an unprecedented radio flare (an increase in flux by a factor of $\sim 60$ over a few months) and the emergence of a spatially resolved jet at $0.1-0.3$ pc scales (Meyer et al. 2024). Companion work has also detected a recurrent quasi-periodic oscillation (QPO) in the $2-10$ keV band with an increasing frequency ($1-2$ mHz) over the same period (Masterson et al., 2025). During this time, the soft X-rays ($0.3-2$ keV) monotonically increased by a factor of $\sim 8$, while the UV emission remained near-steady with $<30\%$ variation and the $2-10$ keV flux showed variation by a factor $\lesssim 2$. The weak variation of the $2-10$ keV X-ray emission and the stability of the UV emission suggest that the magnetic energy density and accretion rate are relatively unchanged, and that the jet could be launched due to a reconfiguration of the magnetic field (toroidal to poloidal) close to the black hole. Advecting poloidal flux onto the event horizon would trigger the Blandford-Znajek (BZ) mechanism, leading to the onset of the jet. The concurrent softening of the coronal slope (from $\Gamma= 2.70\pm 0.04$ to $\Gamma=3.27\pm 0.04$), the appearance of a QPO, and low coronal temperature ($kT_{e}=8_{-3}^{+8}$ keV) during the radio outburst suggest that the poloidal field reconfiguration can significantly impact coronal properties and thus influence jet dynamics. These extraordinary findings in real time are crucial for coronal and jet plasma studies, particularly as our results are independent of coronal geometry., Comment: Submitted to ApJ after minor referee comments

Gastrointestinal (GI) bleeding is a serious medical condition that presents significant diagnostic challenges, particularly in settings with limited access to healthcare resources. Wireless Capsule Endoscopy (WCE) has emerged as a powerful diagnostic tool for visualizing the GI tract, but it requires time-consuming manual analysis by experienced gastroenterologists, which is prone to human error and inefficient given the increasing number of patients.To address this challenge, we propose ClassifyViStA, an AI-based framework designed for the automated detection and classification of bleeding and non-bleeding frames from WCE videos. The model consists of a standard classification path, augmented by two specialized branches: an implicit attention branch and a segmentation branch.The attention branch focuses on the bleeding regions, while the segmentation branch generates accurate segmentation masks, which are used for classification and interpretability. The model is built upon an ensemble of ResNet18 and VGG16 architectures to enhance classification performance. For the bleeding region detection, we implement a Soft Non-Maximum Suppression (Soft NMS) approach with YOLOv8, which improves the handling of overlapping bounding boxes, resulting in more accurate and nuanced detections.The system's interpretability is enhanced by using the segmentation masks to explain the classification results, offering insights into the decision-making process similar to the way a gastroenterologist identifies bleeding regions. Our approach not only automates the detection of GI bleeding but also provides an interpretable solution that can ease the burden on healthcare professionals and improve diagnostic efficiency. Our code is available at ClassifyViStA.

``Little Red Dots'' (LRDs) are an abundant high-redshift population newly discovered by the James Webb Space Telescope (JWST). They are characterized by a red color in the rest-frame optical band, compact morphology, and broad Balmer emission lines (${\rm FWHM} \gtrsim 1000~{\rm km\,s^{-1}}$) that suggest an AGN nature. Using a method of pixel-by-pixel color selection and relaxing the compactness criteria, we identify three of the first dual LRD candidates in the COSMOS-Web survey with projected separations of $0.\!\!^{\prime\prime}2-0.\!\!^{\prime\prime}4$ (1-2 pkpc at their photometric redshifts). A comparison between existing LRD samples and mock data reveals that the projected separations of these dual LRD candidates are unlikely to result from chance projections of objects at different redshifts. In one case (CW-B5-15958), the dual LRD includes two bright sources ($m_{\rm F444W}=24.3$ and $24.8$) with characteristic V-shape spectral energy distribution (SEDs) and photometric redshifts consistent with each other. We find that CW-B5-15958 has a faint off-centered component and a companion galaxy. In the other two dual systems, the brighter LRD exhibits a V-shape SED, while the fainter LRD ($m_{\rm F444W}\gtrsim26$) is undetected in both F115W and F150W. These discoveries suggest that the angular auto-correlation function (ACF) of LRDs exhibits a significant excess ($\sim3\times10^2$ times) on sub-arcsec (kilo-parsec) separations compared to the extrapolation of a power-law ACF of JWST-found AGNs measured over $10^{\prime\prime}-100^{\prime\prime}$. Follow-up spectroscopic confirmation of their redshifts and the construction of a larger sample are essential to advance our understanding of the evolution of supermassive black holes and the importance of mergers in the early universe., Comment: 12 pages, 5 figures, and 1 table. Comments are welcome

The LLM-as-judge paradigm is increasingly being adopted for automated evaluation of model outputs. While LLM judges have shown promise on constrained evaluation tasks, closed source LLMs display critical shortcomings when deployed in real world applications due to challenges of fine grained metrics and explainability, while task specific evaluation models lack cross-domain generalization. We introduce GLIDER, a powerful 3B evaluator LLM that can score any text input and associated context on arbitrary user defined criteria. GLIDER shows higher Pearson's correlation than GPT-4o on FLASK and greatly outperforms prior evaluation models, achieving comparable performance to LLMs 17x its size. GLIDER supports fine-grained scoring, multilingual reasoning, span highlighting and was trained on 685 domains and 183 criteria. Extensive qualitative analysis shows that GLIDER scores are highly correlated with human judgments, with 91.3% human agreement. We have open-sourced GLIDER to facilitate future research.

Active Learning (AL) is a user-interactive approach aimed at reducing annotation costs by selecting the most crucial examples to label. Although AL has been extensively studied for image classification tasks, the specific scenario of interactive image retrieval has received relatively little attention. This scenario presents unique characteristics, including an open-set and class-imbalanced binary classification, starting with very few labeled samples. We introduce a novel batch-mode Active Learning framework named GAL (Greedy Active Learning) that better copes with this application. It incorporates a new acquisition function for sample selection that measures the impact of each unlabeled sample on the classifier. We further embed this strategy in a greedy selection approach, better exploiting the samples within each batch. We evaluate our framework with both linear (SVM) and non-linear MLP/Gaussian Process classifiers. For the Gaussian Process case, we show a theoretical guarantee on the greedy approximation. Finally, we assess our performance for the interactive content-based image retrieval task on several benchmarks and demonstrate its superiority over existing approaches and common baselines. Code is available at https://github.com/barleah/GreedyAL., Comment: Accepted to Transactions on Machine Learning Research (TMLR)

Image restoration aims to recover high-quality images from degraded observations. When the degradation process is known, the recovery problem can be formulated as an inverse problem, and in a Bayesian context, the goal is to sample a clean reconstruction given the degraded observation. Recently, modern pretrained diffusion models have been used for image restoration by modifying their sampling procedure to account for the degradation process. However, these methods often rely on certain approximations that can lead to significant errors and compromised sample quality. In this paper, we provide the first rigorous analysis of this approximation error for linear inverse problems under distributional assumptions on the space of natural images, demonstrating cases where previous works can fail dramatically. Motivated by our theoretical insights, we propose a simple modification to existing diffusion-based restoration methods. Our approach introduces a time-varying low-pass filter in the frequency domain of the measurements, progressively incorporating higher frequencies during the restoration process. We develop an adaptive curriculum for this frequency schedule based on the underlying data distribution. Our method significantly improves performance on challenging image restoration tasks including motion deblurring and image dehazing.

AI workloads, particularly those driven by deep learning, are introducing novel usage patterns to high-performance computing (HPC) systems that are not comprehensively captured by standard HPC benchmarks. As one of the largest academic research centers dedicated to deep learning, Mila identified the need to develop a custom benchmarking suite to address the diverse requirements of its community, which consists of over 1,000 researchers. This report introduces Milabench, the resulting benchmarking suite. Its design was informed by an extensive literature review encompassing 867 papers, as well as surveys conducted with Mila researchers. This rigorous process led to the selection of 26 primary benchmarks tailored for procurement evaluations, alongside 16 optional benchmarks for in-depth analysis. We detail the design methodology, the structure of the benchmarking suite, and provide performance evaluations using GPUs from NVIDIA, AMD, and Intel. The Milabench suite is open source and can be accessed at github.com/mila-iqia/milabench.

Deep neural networks (DNNS) excel at learning from static datasets but struggle with continual learning, where data arrives sequentially. Catastrophic forgetting, the phenomenon of forgetting previously learned knowledge, is a primary challenge. This paper introduces EXponentially Averaged Class-wise Feature Significance (EXACFS) to mitigate this issue in the class incremental learning (CIL) setting. By estimating the significance of model features for each learned class using loss gradients, gradually aging the significance through the incremental tasks and preserving the significant features through a distillation loss, EXACFS effectively balances remembering old knowledge (stability) and learning new knowledge (plasticity). Extensive experiments on CIFAR-100 and ImageNet-100 demonstrate EXACFS's superior performance in preserving stability while acquiring plasticity.

Token generation speed is critical to power the next wave of AI inference applications. GPUs significantly underperform during token generation due to synchronization overheads at kernel boundaries, utilizing only 21% of their peak memory bandwidth. While recent dataflow architectures mitigate these overheads by enabling aggressive fusion of decoder layers into a single kernel, they too leave performance on the table due to synchronization penalties at layer boundaries. This paper presents kernel looping, a specialized global optimization technique which exploits an optimization opportunity brought by combining the unique layer-level fusion possible in modern dataflow architectures with the repeated layer structure found in language models. Kernel looping eliminates synchronization costs between consecutive calls to the same kernel by transforming these calls into a single call to a modified kernel containing a pipelined outer loop. We evaluate kernel looping on the SambaNova SN40L Reconfigurable Dataflow Unit (RDU), a commercial dataflow accelerator for AI. Experiments demonstrate that kernel looping speeds up the decode phase of a wide array of powerful open-source models by up to 2.2$\times$ on SN40L. Kernel looping allows scaling of decode performance over multiple SN40L sockets, achieving speedups of up to 2.5$\times$. Finally, kernel looping enables SN40L to achieve over 90% of peak performance on 8 and 16 sockets and achieve a speedup of up to 3.7$\times$ over DGX H100. Kernel looping, as well as the models evaluated in this paper, are deployed in production in a commercial AI inference cloud.

Prime-Cam, one of the primary instruments for the Fred Young Submillimeter Telescope (FYST) developed by the CCAT Collaboration, will house up to seven instrument modules, with the first operating at 280 GHz. Each module will include three arrays of superconducting microwave kinetic inductance detectors (KIDs). The first KID array fabricated for the 280 GHz module uses titanium-nitride (TiN) as the superconducting material and has 3,456 individual detectors, while the other two arrays use aluminum. This paper presents the design and laboratory characterization of the 280 GHz TiN array, which is cooled below its critical temperature to ~0.1 K and read out over six RF feedlines. LED mapping, a technique for matching the measured resonant frequency of a detector to its physical position, was performed on the array so that the results can be used to lithographically trim the KID capacitors and increase the yield of the array by reducing frequency collisions. We present the methods and results of LED mapping the 280 GHz TiN KID array before deployment on FYST., Comment: 4 pages, 4 figures, submitted to IEEE Transactions on Applied Superconductivity (IEEE TAS)

Activation functions introduce non-linearity into Neural Networks, enabling them to learn complex patterns. Different activation functions vary in speed and accuracy, ranging from faster but less accurate options like ReLU to slower but more accurate functions like SiLU or SELU. Typically, same activation function is used throughout an entire model architecture. In this paper, we conduct a comprehensive study on the effects of using mixed activation functions in YOLO-based models, evaluating their impact on latency, memory usage, and accuracy across CPU, NPU, and GPU edge devices. We also propose a novel approach that leverages Neural Architecture Search (NAS) to design YOLO models with optimized mixed activation functions.The best model generated through this method demonstrates a slight improvement in mean Average Precision (mAP) compared to baseline model (SiLU), while it is 22.28% faster and consumes 64.15% less memory on the reference NPU device., Comment: 7 pages, 4 figures, FITML workshop, NeuRIPS 2024

This paper presents a theoretical framework for an AI-driven data quality monitoring system designed to address the challenges of maintaining data quality in high-volume environments. We examine the limitations of traditional methods in managing the scale, velocity, and variety of big data and propose a conceptual approach leveraging advanced machine learning techniques. Our framework outlines a system architecture that incorporates anomaly detection, classification, and predictive analytics for real-time, scalable data quality management. Key components include an intelligent data ingestion layer, adaptive preprocessing mechanisms, context-aware feature extraction, and AI-based quality assessment modules. A continuous learning paradigm is central to our framework, ensuring adaptability to evolving data patterns and quality requirements. We also address implications for scalability, privacy, and integration within existing data ecosystems. While practical results are not provided, it lays a robust theoretical foundation for future research and implementations, advancing data quality management and encouraging the exploration of AI-driven solutions in dynamic environments.

Dark-field X-ray microscopy (DFXM) is a novel X-ray imaging technique developed at synchrotrons to image along the diffracted beam with a real space resolution of ~100 nm and reciprocal space resolution of $10^{-4}$. Recent implementations of DFXM at X-ray free electron lasers (XFELs) have demonstrated DFXM's ability to visualize the real-time evolution of coherent GHz phonons produced by ultrafast laser excitation of metal transducers. Combining this with DFXM's ability to visualize strain fields due to dislocations makes it possible to study the interaction of GHz coherent phonons with the strain fields of dislocations, along with studying the damping of coherent phonons due to interactions with thermal phonons. For quantitative analysis of phonon-dislocation interactions and phonon damping, a formalism is required to relate phonon dynamics to the strains measured by DFXM. In this work, we use kinematic diffraction theory to simulate DFXM images of the specific coherent phonons in diamond that are generated by the ultrafast laser excitation of a metal transducer. We extend this formalism to also describe imaging of incoherent phonons of sufficiently high frequency, which are relevant for thermal transport, offering future opportunities for DFXM to image signals produced by thermal diffuse scattering. For both coherent and incoherent phonons, we discuss the optimal sampling of real space, reciprocal space and time, and the opportunities offered by the advances in DFXM optics.

Deep reinforcement learning (DRL) has shown success in diverse domains such as robotics, computer games, and recommendation systems. However, like any other software system, DRL-based software systems are susceptible to faults that pose unique challenges for debugging and diagnosing. These faults often result in unexpected behavior without explicit failures and error messages, making debugging difficult and time-consuming. Therefore, automating the monitoring and diagnosis of DRL systems is crucial to alleviate the burden on developers. In this paper, we propose RLExplorer, the first fault diagnosis approach for DRL-based software systems. RLExplorer automatically monitors training traces and runs diagnosis routines based on properties of the DRL learning dynamics to detect the occurrence of DRL-specific faults. It then logs the results of these diagnoses as warnings that cover theoretical concepts, recommended practices, and potential solutions to the identified faults. We conducted two sets of evaluations to assess RLExplorer. Our first evaluation of faulty DRL samples from Stack Overflow revealed that our approach can effectively diagnose real faults in 83% of the cases. Our second evaluation of RLExplorer with 15 DRL experts/developers showed that (1) RLExplorer could identify 3.6 times more defects than manual debugging and (2) RLExplorer is easily integrated into DRL applications., Comment: Accepted for publication in The International Conference on Software Maintenance and Evolution (ICSME 2024)

We explore the relations between the gas-phase metallicity radial profiles (few hundred inner parsec) and multiple galaxy properties for 15 Seyfert galaxies from the AGNIFS (Active Galactic Nuclei Integral Field Spectroscopy) sample using optical Integral Field Unit (IFU) observations from Gemini Multi-Object Spectrographs (GMOS) and Multi Unit Spectroscopic Explorer (MUSE) processed archival data. The data were selected at $z \lesssim 0.013$ within black hole mass range $\left[6<\log \left(M_{\rm BH}/{\rm M_\odot} \right)<9\right]$ with moderate 14--150\,keV X-ray luminosities $\left[42\,\lesssim\,\log L_X (\rm erg\,s^{-1})\,\lesssim\,44\right]$. We estimated the gas-phase metallicity using the strong-line methods and found mean values for the oxygen dependent ($Z \sim 0.75Z_\odot$) and nitrogen dependent ($Z \sim 1.14Z_\odot$) calibrations. These estimates show excellent agreement with $\Delta Z \approx 0.19$ dex and $\Delta Z \approx 0.18$ dex between the mean values from the two strong-line calibrations for GMOS and MUSE respectively, consistent with the order of metallicity uncertainty via the strong-line methods. We contend that our findings align with a scenario wherein local Seyferts have undergone seamless gas accretion histories, resulting in positive metallicity profile over an extended period of time, thereby providing insights into galaxy evolution and the chemical enrichment or depletion of the universe. Additionally, we argue that metal-poor gas inflow from the local interstellar medium (ISM) and accreted through the circumgalactic medium (CGM) onto the galaxy systems regulates the star formation processes by diluting their central metallicity and inverting their metallicity gradients, producing a more prominent anti-correlation between gas-phase metallicity and Eddington ratio., Comment: Accepted for publication in MNRAS; doi: 10.1093/mnras/stae2263

Predictive modeling and time-pattern analysis are increasingly critical in this swiftly shifting retail environment to improve operational efficiency and informed decision-making. This paper reports a comprehensive application of state-of-the-art machine learning to the retailing domain with a specific focus on association rule mining, sequential pattern mining, and time-series forecasting. Association rules: Relationship Mining This provides the key product relationships and customer buying patterns that form the basis of individually tailored marketing campaigns. Sequential pattern mining: Using the PrefixSpan algorithm, it identifies frequent sequences of purchasing products-extremely powerful insights into consumer behavior and also better management of the inventories. What is applied for sales trend forecasting models Prophet applies on historical transaction data over seasonality, holidays, and long-term growth. The forecast results allow predicting demand variations, thus helping in proper inventory alignment and avoiding overstocking or understocking of inventory. Our results are checked through the help of metrics like MAE (Mean Absolute Error) and RMSE (Root Mean Squared Error) to ensure our predictions are strong and accurate. We will combine the aspects of all of these techniques to prove how predictive modeling and temporal pattern analysis can help optimize control over inventory, enhance marketing effectiveness, and position retail businesses as they rise to ever greater heights. This entire methodology demonstrates the flexibility with which data-driven strategies can be leveraged to revitalize traditional retailing practices., Comment: 5 pages, 8 fiqures, published in journal IJCSIS(International Journal of Computer Science and Information Security)

We introduce MCUBench, a benchmark featuring over 100 YOLO-based object detection models evaluated on the VOC dataset across seven different MCUs. This benchmark provides detailed data on average precision, latency, RAM, and Flash usage for various input resolutions and YOLO-based one-stage detectors. By conducting a controlled comparison with a fixed training pipeline, we collect comprehensive performance metrics. Our Pareto-optimal analysis shows that integrating modern detection heads and training techniques allows various YOLO architectures, including legacy models like YOLOv3, to achieve a highly efficient tradeoff between mean Average Precision (mAP) and latency. MCUBench serves as a valuable tool for benchmarking the MCU performance of contemporary object detectors and aids in model selection based on specific constraints., Comment: Code and data are available at https://github.com/Deeplite/deeplite-torch-zoo

We present Gemini near-infrared integral field spectrograph (NIFS) K-band observations of the central 400 pc of NGC 1266, a nearby (D$\approx$30 Mpc) post-starburst galaxy with a powerful multi-phase outflow and a shocked ISM. We detect 7 H$_2$ ro-vibrational emission lines excited thermally to $T$$\sim$2000 K, and weak Br$\gamma$ emission, consistent with a fast C-shock. With these bright H$_2$ lines, we observe the spatial structure of the shock with an unambiguous tracer for the first time. The Br$\gamma$ emission is concentrated in the central $\lesssim$100 pc, indicating that any remaining star-formation in NGC 1266 is in the nucleus while the surrounding cold molecular gas has little on-going star-formation. Though it is unclear what fraction of this Br$\gamma$ emission is from star-formation or the AGN, assuming it is entirely due to star-formation we measure an instantaneous star-formation rate of 0.7 M$_\odot$ yr$^{-1}$, though the star-formation rate may be significantly higher in the presence of additional extinction. NGC 1266 provides a unique laboratory to study the complex interactions between AGN, outflows, shocks, and star-formation, all of which are necessary to unravel the evolution of the post-starburst phase., Comment: ApJ accepted

We present and analyze a control variate strategy based on couplings to reduce the variance of finite difference estimators of sensitivity coefficients, called transport coefficients in the physics literature. We study the bias and variance of a sticky-coupling and a synchronous-coupling based estimator as the finite difference parameter $\eta$ goes to zero. For diffusions with elliptic additive noise, we show that when the drift is contractive outside a compact the bias of a sticky-coupling based estimator is bounded as $\eta \to 0$ and its variance behaves like $\eta^{-1}$, compared to the standard estimator whose bias and variance behave like $\eta^{-1}$ and $\eta^{-2}$, respectively. Under the stronger assumption that the drift is contractive everywhere, we additionally show that the bias and variance of the synchronous-coupling based estimator are both bounded as $\eta \to 0$. Our hypotheses include overdamped Langevin dynamics with many physically relevant non-convex potentials. We illustrate our theoretical results with numerical examples, including overdamped Langevin dynamics with a highly non-convex Lennard-Jones potential to demonstrate both failure of synchronous coupling and the effectiveness of sticky coupling in the not globally contractive setting., Comment: 57 pages, 11 figures

Visual Question Answering (VQA) with multiple choice questions enables a vision-centric evaluation of Multimodal Large Language Models (MLLMs). Although it reliably checks the existence of specific visual abilities, it is easier for the model to select an answer from multiple choices (VQA evaluation) than to generate the answer itself. In this work, we offer a novel perspective: we evaluate how well an MLLM understands a specific visual concept by its ability to uniquely describe two extremely similar images that differ only in the targeted visual concept. Specifically, we assess the ability of MLLMs to capture specific points of visual differences using self-retrieval, i.e., by retrieving the target image using its generated caption against the other image in the pair serving as the distractor. We curate 247 highly similar image pairs as part of the D3 benchmark. For each image pair, the model is prompted to: (1) Detect a specific visual difference, and (2) Describe the target image uniquely such that it (3) Discriminates the target image from the distractor. Self-retrieval within D3 enables whitebox evaluation across six different visual patterns, revealing that current models struggle to independently discern fine-grained visual differences, with open-source models failing to outperform random guess., Comment: ECCV 2024 Workshop EVAL-FoMo; Project Page: https://katha-ai.github.io/projects/detect-describe-discriminate/

Abnormal event detection or anomaly detection in surveillance videos is currently a challenge because of the diversity of possible events. Due to the lack of anomalous events at training time, anomaly detection requires the design of learning methods without supervision. In this work we propose an unsupervised approach for video anomaly detection with the aim to jointly optimize the objectives of the deep neural network and the anomaly detection task using a hybrid architecture. Initially, a convolutional autoencoder is pre-trained in an unsupervised manner with a fusion of depth, motion and appearance features. In the second step, we utilize the encoder part of the pre-trained autoencoder and extract the embeddings of the fused input. Now, we jointly train/ fine tune the encoder to map the embeddings to a hypercenter. Thus, embeddings of normal data fall near the hypercenter, whereas embeddings of anomalous data fall far away from the hypercenter.

Image captioning systems are unable to generate fine-grained captions as they are trained on data that is either noisy (alt-text) or generic (human annotations). This is further exacerbated by maximum likelihood training that encourages generation of frequently occurring phrases. Previous works have tried to address this limitation by fine-tuning captioners with a self-retrieval (SR) reward. However, we find that SR fine-tuning has a tendency to reduce caption faithfulness and even hallucinate. In this work, we circumvent this bottleneck by improving the MLE initialization of the captioning system and designing a curriculum for the SR fine-tuning process. To this extent, we present (1) Visual Caption Boosting, a novel framework to instill fine-grainedness in generic image captioning datasets while remaining anchored in human annotations; and (2) BagCurri, a carefully designed training curriculum that more optimally leverages the contrastive nature of the self-retrieval reward. Jointly, they enable the captioner to describe fine-grained aspects in the image while preserving faithfulness to ground-truth captions. Our approach outperforms previous work by +8.9% on SR against 99 random distractors (RD100) (Dessi et al., 2023); and +7.6% on ImageCoDe. Additionally, existing metrics to evaluate captioning systems fail to reward diversity or evaluate a model's fine-grained understanding ability. Our third contribution addresses this by proposing self-retrieval from the lens of evaluation. We introduce TrueMatch, a benchmark comprising bags of highly similar images that uses SR to assess the captioner's ability to capture subtle visual distinctions. We evaluate and compare several state-of-the-art open-source MLLMs on TrueMatch, and find that our SR approach outperforms them all by a significant margin (e.g. +4.8% - 7.1% over Cambrian) while having 1-2 orders of magnitude fewer parameters.

Prime-Cam, a first-generation science instrument for the Atacama-based Fred Young Submillimeter Telescope, is being built by the CCAT Collaboration to observe at millimeter and submillimeter wavelengths using kinetic inductance detectors (KIDs). Prime-Cam's 280 GHz instrument module will deploy with two aluminum-based KID arrays and one titanium nitride-based KID array, totaling approximately 10,000 detectors at the focal plane, all of which have been fabricated and are currently undergoing testing. One complication of fielding large arrays of KIDs under dynamic loading conditions is tuning the detector tone powers to maximize signal-to-noise while avoiding bifurcation due to the nonlinear kinetic inductance. For aluminum-based KIDs, this is further complicated by additional nonlinear effects which couple tone power to resonator quality factors and resonant frequencies. While both nonequilibrium quasiparticle dynamics and two-level system fluctuations have been shown to give rise to qualitatively similar distortions, modeling these effects alongside nonlinear kinetic inductance is inefficient when fitting thousands of resonators on-sky with existing models. For this reason, it is necessary to have a detailed understanding of the nonlinear effects across relevant detector loading conditions, including how they impact on on-sky noise and how to diagnose the detector's relative performance. We present a study of the competing nonlinearities seen in Prime-Cam's 280 GHz aluminum KIDs, with a particular emphasis on the resulting distortions to the resonator line shape and how these impact detector parameter estimation., Comment: 10 pages, 4 figures, Conference proceedings from SPIE Astronomical Telescopes + Instrumentation (AS24)

Harnessing pre-trained LLMs to improve ASR systems, particularly for low-resource languages, is now an emerging area of research. Existing methods range from using LLMs for ASR error correction to tightly coupled systems that replace the ASR decoder with the LLM. These approaches either increase decoding time or require expensive training of the cross-attention layers. We propose SALSA, which couples the decoder layers of the ASR to the LLM decoder, while synchronously advancing both decoders. Such coupling is performed with a simple projection of the last decoder state, and is thus significantly more training efficient than earlier approaches. A challenge of our proposed coupling is handling the mismatch between the tokenizers of the LLM and ASR systems. We handle this mismatch using cascading tokenization with respect to the LLM and ASR vocabularies. We evaluate SALSA on 8 low-resource languages in the FLEURS benchmark, yielding substantial WER reductions of up to 38%., Comment: Accepted to INTERSPEECH 2024

We introduce Contested Logistics Games, a variant of logistics problems that account for the presence of an adversary that can disrupt the movement of goods in selected areas. We model this as a large two-player zero-sum one-shot game played on a graph representation of the physical world, with the optimal logistics plans described by the (possibly randomized) Nash equilibria of this game. Our logistics model is fairly sophisticated, and is able to handle multiple modes of transport and goods, accounting for possible storage of goods in warehouses, as well as Leontief utilities based on demand satisfied. We prove computational hardness results related to equilibrium finding and propose a practical double-oracle solver based on solving a series of best-response mixed-integer linear programs. We experiment on both synthetic and real-world maps, demonstrating that our proposed method scales to reasonably large games. We also demonstrate the importance of explicitly modeling the capabilities of the adversary via ablation studies and comparisons with a naive logistics plan based on heuristics., Comment: GameSec '24

This paper introduces the new and powerful image patch descriptor based on second order image statistics/derivatives. Here, the image patch is treated as a 3D surface with intensity being the 3rd dimension. The considered 3D surface has a rich set of second order features/statistics such as ridges, valleys, cliffs and so on, that can be easily captured by using the difference of rotating semi Gaussian filters. The originality of this method is based on successfully combining the response of the directional filters with that of the Difference of Gaussian (DOG) approach. The obtained descriptor shows a good discriminative power when dealing with the variations in illumination, scale, rotation, blur, viewpoint and compression. The experiments on image matching, demonstrates the advantage of the obtained descriptor when compared to its first order counterparts such as SIFT, DAISY, GLOH, GIST and LIDRIC.

Given an irreducible representation of $SL_2(F_q)$ for an odd prime $q\geq 5$, we find the dimension of the space of cusp forms with respect to the full modular group taking values in the representation space. The dimension equals the multiplicity of the representation in the space of classical cusp forms with respect to the principal congruence subgroup of level $q$., Comment: 10 pages, any comments and suggestions welcomes

The main result of this article proves the nonvanishing of cuspidal cohomology for $GL(n)$ over a number field which is Galois over its maximal totally real subfield. The proof uses the internal structure of a strongly-pure weight that can possibly support cuspidal cohomology and the foundational work of Borel, Labesse, and Schwermer.

Speech accents present a serious challenge to the performance of state-of-the-art end-to-end Automatic Speech Recognition (ASR) systems. Even with self-supervised learning and pre-training of ASR models, accent invariance is seldom achieved. In this work, we propose an accent-aware adaptation technique for self-supervised learning that introduces a trainable set of accent-specific codebooks to the self-supervised architecture. These learnable codebooks enable the model to capture accent specific information during pre-training, that is further refined during ASR finetuning. On the Mozilla Common Voice dataset, our proposed approach outperforms all other accent-adaptation approaches on both seen and unseen English accents, with up to 9% relative reduction in word error rate (WER)., Comment: Accepted to INTERSPEECH 2024

Convolutions have become essential in state-of-the-art end-to-end Automatic Speech Recognition~(ASR) systems due to their efficient modelling of local context. Notably, its use in Conformers has led to superior performance compared to vanilla Transformer-based ASR systems. While components other than the convolution module in the Conformer have been reexamined, altering the convolution module itself has been far less explored. Towards this, we introduce Multi-Convformer that uses multiple convolution kernels within the convolution module of the Conformer in conjunction with gating. This helps in improved modeling of local dependencies at varying granularities. Our model rivals existing Conformer variants such as CgMLP and E-Branchformer in performance, while being more parameter efficient. We empirically compare our approach with Conformer and its variants across four different datasets and three different modelling paradigms and show up to 8% relative word error rate~(WER) improvements., Comment: Accepted to INTERSPEECH 2024

Long gamma-ray bursts (GRBs) offer significant insights into cosmology due to their high energy emissions and the potential to probe the early universe. The Amati relation, which links the intrinsic peak energy to the isotropic energy, is crucial for understanding their cosmological applications. This study investigates the redshift-driven heterogeneity of the Amati correlation in long GRBs. Analyzing 221 long GRBs with redshifts from 0.034 to 8.2 we divided the dataset based on redshift thresholds of 1.5 and 2. Using Bayesian marginalization and Reichart's likelihood approach, we found significant differences in the Amati parameters between low and high redshift subgroups. These variations, differing by approximately $2\sigma$ at $z = 1.5$ and more than $1\sigma$ at $z = 2$, suggest an evolution in the GRB population with redshift, possibly reflecting changes in host galaxy properties. However, selection effects and instrumental biases may also contribute. Our results challenge the assumption of the Amati relation's universality and underscore the need for larger datasets and more precise measurements from upcoming missions like Transient High-Energy Sky and Early Universe Surveyor (THESEUS) and enhanced X-ray Timing and Polarimetry mission (eXTP) to refine our understanding of GRB physics., Comment: 21 pages, 6 figures

Compositionality is a fundamental aspect of vision-language understanding and is especially required for videos since they contain multiple entities (e.g. persons, actions, and scenes) interacting dynamically over time. Existing benchmarks focus primarily on perception capabilities. However, they do not study binding, the ability of a model to associate entities through appropriate relationships. To this end, we propose VELOCITI, a new benchmark building on complex movie clips and dense semantic role label annotations to test perception and binding in video language models (contrastive and Video-LLMs). Our perception-based tests require discriminating video-caption pairs that share similar entities, and the binding tests require models to associate the correct entity to a given situation while ignoring the different yet plausible entities that also appear in the same video. While current state-of-the-art models perform moderately well on perception tests, accuracy is near random when both entities are present in the same video, indicating that they fail at binding tests. Even the powerful Gemini 1.5 Flash has a substantial gap (16-28%) with respect to human accuracy in such binding tests., Comment: 26 pages, 17 figures, 3 tables

Language Models have previously shown strong negotiation capabilities in closed domains where the negotiation strategy prediction scope is constrained to a specific setup. In this paper, we first show that these models are not generalizable beyond their original training domain despite their wide-scale pretraining. Following this, we propose an automated framework called GNOME, which processes existing human-annotated, closed-domain datasets using Large Language Models and produces synthetic open-domain dialogues for negotiation. GNOME improves the generalizability of negotiation systems while reducing the expensive and subjective task of manual data curation. Through our experimental setup, we create a benchmark comparing encoder and decoder models trained on existing datasets against datasets created through GNOME. Our results show that models trained on our dataset not only perform better than previous state of the art models on domain specific strategy prediction, but also generalize better to previously unseen domains.

We present the results of an investigation of a highly variable CIV broad absorption-line feature in the quasar SBS 1408+544 (z=2.337) that shows a significant shift in velocity over time. This source was observed as a part of the Sloan Digital Sky Survey Reverberation Mapping Project and the SDSS-V Black Hole Mapper Reverberation Mapping Project, and has been included in two previous studies, both of which identified significant variability in a high-velocity CIV broad absorption line (BAL) on timescales of just a few days in the quasar rest frame. Using ~130 spectra acquired over eight years of spectroscopic monitoring with SDSS, we have determined that this BAL is not only varying in strength, but is also systematically shifting to higher velocities. Using cross-correlation methods, we measure the velocity shifts (and corresponding acceleration) of the BAL on a wide range of timescales, measuring an overall velocity shift of delta v = -683 (+89, -84) km s-1 over the 8-year monitoring period. This corresponds to an average rest-frame acceleration of a=1.04 (+0.14, -0.13) cm s-2, though the magnitude of the acceleration on shorter timescales is not constant throughout. We place our measurements in the context of BAL-acceleration models and examine various possible causes of the observed velocity shift., Comment: Published in the Astrophysical Journal

Large Language Models (LLMs) are being used for a wide variety of tasks. While they are capable of generating human-like responses, they can also produce undesirable output including potentially harmful information, racist or sexist language, and hallucinations. Alignment methods are designed to reduce such undesirable outputs via techniques such as fine-tuning, prompt engineering, and representation engineering. However, existing methods face several challenges: some require costly fine-tuning for every alignment task; some do not adequately remove undesirable concepts, failing alignment; some remove benign concepts, lowering the linguistic capabilities of LLMs. To address these issues, we propose Parsimonious Concept Engineering (PaCE), a novel activation engineering framework for alignment. First, to sufficiently model the concepts, we construct a large-scale concept dictionary in the activation space, in which each atom corresponds to a semantic concept. Given any alignment task, we instruct a concept partitioner to efficiently annotate the concepts as benign or undesirable. Then, at inference time, we decompose the LLM activations along the concept dictionary via sparse coding, to accurately represent the activations as linear combinations of benign and undesirable components. By removing the latter ones from the activations, we reorient the behavior of the LLM towards the alignment goal. We conduct experiments on tasks such as response detoxification, faithfulness enhancement, and sentiment revising, and show that PaCE achieves state-of-the-art alignment performance while maintaining linguistic capabilities., Comment: Accepted in NeurIPS 2024. GitHub repository at https://github.com/peterljq/Parsimonious-Concept-Engineering

Predicting the behavior of AI-driven agents is particularly challenging without a preexisting model. In our paper, we address this by treating AI agents as nonlinear dynamical systems and adopting a probabilistic perspective to predict their statistical behavior using the Perron-Frobenius (PF) operator. We formulate the approximation of the PF operator as an entropy minimization problem, which can be solved by leveraging the Markovian property of the operator and decomposing its spectrum. Our data-driven methodology simultaneously approximates the PF operator to perform prediction of the evolution of the agents and also predicts the terminal probability density of AI agents, such as robotic systems and generative models. We demonstrate the effectiveness of our prediction model through extensive experiments on practical systems driven by AI algorithms., Comment: 12 pages, 4 figures, conference

The task of Visual Place Recognition (VPR) is to predict the location of a query image from a database of geo-tagged images. Recent studies in VPR have highlighted the significant advantage of employing pre-trained foundation models like DINOv2 for the VPR task. However, these models are often deemed inadequate for VPR without further fine-tuning on task-specific data. In this paper, we propose a simple yet powerful approach to better exploit the potential of a foundation model for VPR. We first demonstrate that features extracted from self-attention layers can serve as a powerful re-ranker for VPR. Utilizing these features in a zero-shot manner, our method surpasses previous zero-shot methods and achieves competitive results compared to supervised methods across multiple datasets. Subsequently, we demonstrate that a single-stage method leveraging internal ViT layers for pooling can generate global features that achieve state-of-the-art results, even when reduced to a dimensionality as low as 128D. Nevertheless, incorporating our local foundation features for re-ranking, expands this gap. Our approach further demonstrates remarkable robustness and generalization, achieving state-of-the-art results, with a significant gap, in challenging scenarios, involving occlusion, day-night variations, and seasonal changes.

Personalized retrieval and segmentation aim to locate specific instances within a dataset based on an input image and a short description of the reference instance. While supervised methods are effective, they require extensive labeled data for training. Recently, self-supervised foundation models have been introduced to these tasks showing comparable results to supervised methods. However, a significant flaw in these models is evident: they struggle to locate a desired instance when other instances within the same class are presented. In this paper, we explore text-to-image diffusion models for these tasks. Specifically, we propose a novel approach called PDM for Personalized Features Diffusion Matching, that leverages intermediate features of pre-trained text-to-image models for personalization tasks without any additional training. PDM demonstrates superior performance on popular retrieval and segmentation benchmarks, outperforming even supervised methods. We also highlight notable shortcomings in current instance and segmentation datasets and propose new benchmarks for these tasks., Comment: Accepted to NeurIPS 2024

Monolithic large language models (LLMs) like GPT-4 have paved the way for modern generative AI applications. Training, serving, and maintaining monolithic LLMs at scale, however, remains prohibitively expensive and challenging. The disproportionate increase in compute-to-memory ratio of modern AI accelerators have created a memory wall, necessitating new methods to deploy AI. Composition of Experts (CoE) is an alternative modular approach that lowers the cost and complexity of training and serving. However, this approach presents two key challenges when using conventional hardware: (1) without fused operations, smaller models have lower operational intensity, which makes high utilization more challenging to achieve; and (2) hosting a large number of models can be either prohibitively expensive or slow when dynamically switching between them. In this paper, we describe how combining CoE, streaming dataflow, and a three-tier memory system scales the AI memory wall. We describe Samba-CoE, a CoE system with 150 experts and a trillion total parameters. We deploy Samba-CoE on the SambaNova SN40L Reconfigurable Dataflow Unit (RDU) - a commercial dataflow accelerator architecture that has been co-designed for enterprise inference and training applications. The chip introduces a new three-tier memory system with on-chip distributed SRAM, on-package HBM, and off-package DDR DRAM. A dedicated inter-RDU network enables scaling up and out over multiple sockets. We demonstrate speedups ranging from 2$\times$ to 13$\times$ on various benchmarks running on eight RDU sockets compared with an unfused baseline. We show that for CoE inference deployments, the 8-socket RDU Node reduces machine footprint by up to 19$\times$, speeds up model switching time by 15$\times$ to 31$\times$, and achieves an overall speedup of 3.7$\times$ over a DGX H100 and 6.6$\times$ over a DGX A100., Comment: 2024 57th IEEE/ACM International Symposium on Microarchitecture (MICRO)

In the real world, the strong episode resetting mechanisms that are needed to train agents in simulation are unavailable. The \textit{resetting} assumption limits the potential of reinforcement learning in the real world, as providing resets to an agent usually requires the creation of additional handcrafted mechanisms or human interventions. Recent work aims to train agents (\textit{forward}) with learned resets by constructing a second (\textit{backward}) agent that returns the forward agent to the initial state. We find that the termination and timing of the transitions between these two agents are crucial for algorithm success. With this in mind, we create a new algorithm, Reset Free RL with Intelligently Switching Controller (RISC) which intelligently switches between the two agents based on the agent's confidence in achieving its current goal. Our new method achieves state-of-the-art performance on several challenging environments for reset-free RL., Comment: Published at ICLR 2024

Thermometry based on magnetic resonance has been extensively studied due to its important application in biomedical imaging. In our previous work, we showed that the spin-spin relaxation time (T2) of nuclear magnetic resonance (NMR) in water is a highly sensitive thermometer as T2 scales with the highly temperature-sensitive self-diffusion constant of water. In this work, in addition to temperature dependent self-diffusion constant of a fluid, we utilize the temperature dependent magnetization of 4 nm SPIONs to improve T2 sensitivity (4.96) by 1.4 times over self-diffusion (3.48) alone in hexane between 248 K and 333 K. To extend the application of NMR T2 thermometry to engineering systems, we also investigate the temperature dependence of T2 in mineral oil (Thermo Scientific, J62592), which exhibits remarkably high sensitivity (11.62) between 273 K and 353 K. This result implies that applications of NMR T2 thermometry in heat transfer fluids are promising. NMR thermometry, however, is generally not applicable to solids. Therefore, we also evaluate the potential of electron spin resonance (ESR) thermometry with SPIONs in solids between 100 K and 290 K, for potential temperature monitoring in biomedical and engineering applications. The size and concentration effects on ESR signals are studied systematically, and our results show that the temperature dependent linewidth follows a T^-2 law for 4 nm SPIONs, while the concentration of SPIONs has no impact on the temperature dependence of the ESR linewidth. The linewidth at room temperature at 9.4 GHz is 10.5 mT. Combining our NMR and ESR results, we find that to obtain higher temperature sensitivity in a magnetic resonance technique using SPIONs, SPION with a small magnetic moment, i.e., a small volume and reduced magnetization, are beneficial., Comment: 45 pages, 16 figures

Self-supervised learned (SSL) models such as Wav2vec and HuBERT yield state-of-the-art results on speech-related tasks. Given the effectiveness of such models, it is advantageous to use them in conventional ASR systems. While some approaches suggest incorporating these models as a trainable encoder or a learnable frontend, training such systems is extremely slow and requires a lot of computation cycles. In this work, we propose two simple approaches that use (1) framewise addition and (2) cross-attention mechanisms to efficiently incorporate the representations from the SSL model(s) into the ASR architecture, resulting in models that are comparable in size with standard encoder-decoder conformer systems while also avoiding the usage of SSL models during training. Our approach results in faster training and yields significant performance gains on the Librispeech and Tedlium datasets compared to baselines. We further provide detailed analysis and ablation studies that demonstrate the effectiveness of our approach., Comment: Accepted to ENLSP workshop, NeurIPS 2023

Tuberculosis (TB) remains a global health threat, ranking among the leading causes of mortality worldwide. In this context, machine learning (ML) has emerged as a transformative force, providing innovative solutions to the complexities associated with TB treatment.This study explores how machine learning, especially with tabular data, can be used to predict Tuberculosis (TB) treatment outcomes more accurately. It transforms this prediction task into a binary classification problem, generating risk scores from patient data sourced from NIKSHAY, India's national TB control program, which includes over 500,000 patient records. Data preprocessing is a critical component of the study, and the model achieved an recall of 98% and an AUC-ROC score of 0.95 on the validation set, which includes 20,000 patient records.We also explore the use of Natural Language Processing (NLP) for improved model learning. Our results, corroborated by various metrics and ablation studies, validate the effectiveness of our approach. The study concludes by discussing the potential ramifications of our research on TB eradication efforts and proposing potential avenues for future work. This study marks a significant stride in the battle against TB, showcasing the potential of machine learning in healthcare.

Optimizing grain boundary characteristics in polycrystalline materials can improve their properties. Many processing methods have been developed for grain boundary manipulation, including the use of intense radiation in certain applications. In this work, we used X-ray free electron laser pulses to irradiate single-crystalline bismuth selenide (Bi2Se3) and observed grain boundary formation and subsequent grain rotation in response to the X-ray radiation. Our observations with simultaneous transmission X-ray microscopy and X-ray diffraction demonstrate how intense X- ray radiation can rapidly change size and texture of grains., Comment: 37 pages, 11 figures including 6 supplemental figures

The Wide-field Spectroscopic Telescope (WST) is proposed as a new facility dedicated to the efficient delivery of spectroscopic surveys. This white paper summarises the initial concept as well as the corresponding science cases. WST will feature simultaneous operation of a large field-of-view (3 sq. degree), a high multiplex (20,000) multi-object spectrograph (MOS) and a giant 3x3 sq. arcmin integral field spectrograph (IFS). In scientific capability these requirements place WST far ahead of existing and planned facilities. Given the current investment in deep imaging surveys and noting the diagnostic power of spectroscopy, WST will fill a crucial gap in astronomical capability and work synergistically with future ground and space-based facilities. This white paper shows that WST can address outstanding scientific questions in the areas of cosmology; galaxy assembly, evolution, and enrichment, including our own Milky Way; origin of stars and planets; time domain and multi-messenger astrophysics. WST's uniquely rich dataset will deliver unforeseen discoveries in many of these areas. The WST Science Team (already including more than 500 scientists worldwide) is open to the all astronomical community. To register in the WST Science Team please visit https://www.wstelescope.com/for-scientists/participate, Comment: 194 pages, 66 figures. Comments are welcome (wstelescope@gmail.com)

In this paper, we introduce the first algorithmic framework for Blackwell approachability on the sequence-form polytope, the class of convex polytopes capturing the strategies of players in extensive-form games (EFGs). This leads to a new class of regret-minimization algorithms that are stepsize-invariant, in the same sense as the Regret Matching and Regret Matching$^+$ algorithms for the simplex. Our modular framework can be combined with any existing regret minimizer over cones to compute a Nash equilibrium in two-player zero-sum EFGs with perfect recall, through the self-play framework. Leveraging predictive online mirror descent, we introduce Predictive Treeplex Blackwell$^+$ (PTB$^+$), and show a $O(1/\sqrt{T})$ convergence rate to Nash equilibrium in self-play. We then show how to stabilize PTB$^+$ with a stepsize, resulting in an algorithm with a state-of-the-art $O(1/T)$ convergence rate. We provide an extensive set of experiments to compare our framework with several algorithmic benchmarks, including CFR$^+$ and its predictive variant, and we highlight interesting connections between practical performance and the stepsize-dependence or stepsize-invariance properties of classical algorithms.

Dynamic mechanism design is a challenging extension to ordinary mechanism design in which the mechanism designer must make a sequence of decisions over time in the face of possibly untruthful reports of participating agents. Optimizing dynamic mechanisms for welfare is relatively well understood. However, there has been less work on optimizing for other goals (e.g. revenue), and without restrictive assumptions on valuations, it is remarkably challenging to characterize good mechanisms. Instead, we turn to automated mechanism design to find mechanisms with good performance in specific problem instances. In fact, the situation is similar even in static mechanism design. However, in the static case, optimization/machine learning-based automated mechanism design techniques have been successful in finding high-revenue mechanisms in cases beyond the reach of analytical results. We extend the class of affine maximizer mechanisms to MDPs where agents may untruthfully report their rewards. This extension results in a challenging bilevel optimization problem in which the upper problem involves choosing optimal mechanism parameters, and the lower problem involves solving the resulting MDP. Our approach can find truthful dynamic mechanisms that achieve strong performance on goals other than welfare, and can be applied to essentially any problem setting-without restrictions on valuations-for which RL can learn optimal policies., Comment: To be published in the Thirty-Eighth Proceedings of the AAAI Conference on Artificial Intelligence 2024

We propose a novel approach based on Denoising Diffusion Probabilistic Models (DDPMs) to control nonlinear dynamical systems. DDPMs are the state-of-art of generative models that have achieved success in a wide variety of sampling tasks. In our framework, we pose the feedback control problem as a generative task of drawing samples from a target set under control system constraints. The forward process of DDPMs constructs trajectories originating from a target set by adding noise. We learn to control a dynamical system in reverse such that the terminal state belongs to the target set. For control-affine systems without drift, we prove that the control system can exactly track the trajectory of the forward process in reverse, whenever the the Lie bracket based condition for controllability holds. We numerically study our approach on various nonlinear systems and verify our theoretical results. We also conduct numerical experiments for cases beyond our theoretical results on a physics-engine.