Search

Graph neural networks are typically trained on individual datasets, often requiring highly specialized models and extensive hyperparameter tuning. This dataset-specific approach arises because each graph dataset often has unique node features and diverse connectivity structures, making it difficult to build a generalist model. To address these challenges, we introduce a scalable multi-graph multi-task pretraining approach specifically tailored for node classification tasks across diverse graph datasets from different domains. Our method, Graph Foundation Model (GraphFM), leverages a Perceiver-based encoder that employs learned latent tokens to compress domain-specific features into a common latent space. This approach enhances the model's ability to generalize across different graphs and allows for scaling across diverse data. We demonstrate the efficacy of our approach by training a model on 152 different graph datasets comprising over 7.4 million nodes and 189 million edges, establishing the first set of scaling laws for multi-graph pretraining on datasets spanning many domains (e.g., molecules, citation and product graphs). Our results show that pretraining on a diverse array of real and synthetic graphs improves the model's adaptability and stability, while performing competitively with state-of-the-art specialist models. This work illustrates that multi-graph pretraining can significantly reduce the burden imposed by the current graph training paradigm, unlocking new capabilities for the field of graph neural networks by creating a single generalist model that performs competitively across a wide range of datasets and tasks.

We present a comprehensive evaluation of proprietary and open-weights large language models using the first astronomy-specific benchmarking dataset. This dataset comprises 4,425 multiple-choice questions curated from the Annual Review of Astronomy and Astrophysics, covering a broad range of astrophysical topics. Our analysis examines model performance across various astronomical subfields and assesses response calibration, crucial for potential deployment in research environments. Claude-3.5-Sonnet outperforms competitors by up to 4.6 percentage points, achieving 85.0% accuracy. For proprietary models, we observed a universal reduction in cost every 3-to-12 months to achieve similar score in this particular astronomy benchmark. open-weights models have rapidly improved, with LLaMA-3-70b (80.6%) and Qwen-2-72b (77.7%) now competing with some of the best proprietary models. We identify performance variations across topics, with non-English-focused models generally struggling more in exoplanet-related fields, stellar astrophysics, and instrumentation related questions. These challenges likely stem from less abundant training data, limited historical context, and rapid recent developments in these areas. This pattern is observed across both open-weights and proprietary models, with regional dependencies evident, highlighting the impact of training data diversity on model performance in specialized scientific domains. Top-performing models demonstrate well-calibrated confidence, with correlations above 0.9 between confidence and correctness, though they tend to be slightly underconfident. The development for fast, low-cost inference of open-weights models presents new opportunities for affordable deployment in astronomy. The rapid progress observed suggests that LLM-driven research in astronomy may become feasible in the near future., Comment: 45 pages, 12 figures, 7 tables. Published in Astronomy & Computing. AstroMLab homepage: https://astromlab.org/

The current study addresses the control problems posed by a semilinear neutral integro-differential equation with memory. The primary objectives of this study are to investigate the existence of a mild solution and approximate controllability of both linear and semilinear control systems in Banach spaces. To accomplish this, we begin by introducing the concept of a resolvent family associated with the homogeneous neutral integro-differential equation without memory. In the process, we establish some important properties of the resolvent family. Subsequently, we develop approximate controllability results for a linear control problem by constructing a linear-quadratic regulator problem. This involves establishing the existence of an optimal pair and determining the expression of the optimal control that produces the approximate controllability of the linear system. Furthermore, we deduce sufficient conditions for the existence of a mild solution and approximate controllability of a semilinear system in a reflexive Banach space with a uniformly convex dual. Additionally, we delve into the discussion of approximate controllability for a semilinear problem in general Banach space, assuming a Lipschitz type condition on the nonlinear term. Finally, we implement our findings to examine the approximate controllability of certain partial differential equations, demonstrating their practical relevance.

We focus on the problem of Unsupervised Domain Adaptation (\uda) for breast cancer detection from mammograms (BCDM) problem. Recent advancements have shown that masked image modeling serves as a robust pretext task for UDA. However, when applied to cross-domain BCDM, these techniques struggle with breast abnormalities such as masses, asymmetries, and micro-calcifications, in part due to the typically much smaller size of region of interest in comparison to natural images. This often results in more false positives per image (FPI) and significant noise in pseudo-labels typically used to bootstrap such techniques. Recognizing these challenges, we introduce a transformer-based Domain-invariant Mask Annealed Student Teacher autoencoder (D-MASTER) framework. D-MASTER adaptively masks and reconstructs multi-scale feature maps, enhancing the model's ability to capture reliable target domain features. D-MASTER also includes adaptive confidence refinement to filter pseudo-labels, ensuring only high-quality detections are considered. We also provide a bounding box annotated subset of 1000 mammograms from the RSNA Breast Screening Dataset (referred to as RSNA-BSD1K) to support further research in BCDM. We evaluate D-MASTER on multiple BCDM datasets acquired from diverse domains. Experimental results show a significant improvement of 9% and 13% in sensitivity at 0.3 FPI over state-of-the-art UDA techniques on publicly available benchmark INBreast and DDSM datasets respectively. We also report an improvement of 11% and 17% on In-house and RSNA-BSD1K datasets respectively. The source code, pre-trained D-MASTER model, along with RSNA-BSD1K dataset annotations is available at https://dmaster-iitd.github.io/webpage.

We consider non-Gaussian states generated by photon subtraction (PS) and photon addition (PA) on two-mode squeezed coherent (TMSC) states, as resource states for continuous variable (CV) quantum teleportation (QT). To this end, we derive the Wigner characteristic function for the family of photon subtracted and photon added TMSC states, which is then utilized to calculate the fidelity of teleporting a single mode coherent state and a squeezed vacuum state. The analysis shows that while symmetric PS enhances the fidelity of QT in an extensive range of squeezing, asymmetric PS enhances the performance marginally and only in the low squeezing regime. The addition operations on the other hand are less useful, symmetric three-PA leads to a marginal improvement while the other addition operations are useless. We have considered the actual experimental setup for PS and PA operations and computed their success probabilities which should be kept in mind while advocating the use of these operations. We could compute the fidelity of QT for a broad range of states because we analytically derived the Wigner characteristic function for these family of states which we think will be useful for various other applications of these families of states., Comment: Comments are welcome

Recurrent large language models that compete with Transformers in language modeling perplexity are emerging at a rapid rate (e.g., Mamba, RWKV). Excitingly, these architectures use a constant amount of memory during inference. However, due to the limited memory, recurrent LMs cannot recall and use all the information in long contexts leading to brittle in-context learning (ICL) quality. A key challenge for efficient LMs is selecting what information to store versus discard. In this work, we observe the order in which information is shown to the LM impacts the selection difficulty. To formalize this, we show that the hardness of information recall reduces to the hardness of a problem called set disjointness (SD), a quintessential problem in communication complexity that requires a streaming algorithm (e.g., recurrent model) to decide whether inputted sets are disjoint. We empirically and theoretically show that the recurrent memory required to solve SD changes with set order, i.e., whether the smaller set appears first in-context. Our analysis suggests, to mitigate the reliance on data order, we can put information in the right order in-context or process prompts non-causally. Towards that end, we propose: (1) JRT-Prompt, where context gets repeated multiple times in the prompt, effectively showing the model all data orders. This gives $11.0 \pm 1.3$ points of improvement, averaged across $16$ recurrent LMs and the $6$ ICL tasks, with $11.9\times$ higher throughput than FlashAttention-2 for generation prefill (length $32$k, batch size $16$, NVidia H100). We then propose (2) JRT-RNN, which uses non-causal prefix-linear-attention to process prompts and provides $99\%$ of Transformer quality at $360$M params., $30$B tokens and $96\%$ at $1.3$B params., $50$B tokens on average across the tasks, with $19.2\times$ higher throughput for prefill than FA2.

Multi-label radiography image classification has long been a topic of interest in neural networks research. In this paper, we intend to classify such images using convolution neural networks with novel localization techniques. We will use the chest x-ray images to detect thoracic diseases for this purpose. For accurate diagnosis, it is crucial to train the network with good quality images. But many chest X-ray images have irrelevant external objects like distractions created by faulty scans, electronic devices scanned next to lung region, scans inadvertently capturing bodily air etc. To address these, we propose a combination of localization and deep learning algorithms called LeDNet to predict thoracic diseases with higher accuracy. We identify and extract the lung region masks from chest x-ray images through localization. These masks are superimposed on the original X-ray images to create the mask overlay images. DenseNet-121 classification models are then used for feature selection to retrieve features of the entire chest X-ray images and the localized mask overlay images. These features are then used to predict disease classification. Our experiments involve comparing classification results obtained with original CheXpert images and mask overlay images. The comparison is demonstrated through accuracy and loss curve analyses., Comment: 6 pages, 7 figures

In the present work, we extend the $f(Q)$ symmetric teleparallel gravity by introducing an arbitrary coupling between the non-metricity $Q$ and matter Lagrangian $L_m$ in the Lagrangian density $f$ of the theory, which thus leads to the $f\left(Q,L_m\right)$ theory. This generalisation encompasses Coincident General Relativity (CGR), and the Symmetric Teleparallel Equivalent to GR (STEGR). Using the metric formalism, we derive the field equation of the theory, which generalizes the field equations of $f(Q)$ gravity. From the study of the covariant divergence of the field equations, it follows that the presence of the geometry-matter coupling leads to the non-conservation of the matter energy-momentum tensor. The cosmological implications of the theory are investigated in the case of a flat, homogeneous, and isotropic Friedmann-Lemaitre-Robertson-Walker geometry. As a first step in this direction, we obtain the modified Friedmann equations for the $f(Q,L_m)$ gravity in a general form. Specific cosmological models are investigated for several choices of $f(Q,L_m)$, including $f(Q,L_m)=-\alpha Q + 2L_m + \beta$, and $f(Q,L_m)=- \alpha Q + (2L_m)^2 + \beta$, respectively. Comparative analyses with the standard $\Lambda$ CDM paradigm are carried out, and the observational implications of the models are investigated in detail., Comment: Updated version submitted to JCAP

Uncovering latent values and opinions embedded in large language models (LLMs) can help identify biases and mitigate potential harm. Recently, this has been approached by prompting LLMs with survey questions and quantifying the stances in the outputs towards morally and politically charged statements. However, the stances generated by LLMs can vary greatly depending on how they are prompted, and there are many ways to argue for or against a given position. In this work, we propose to address this by analysing a large and robust dataset of 156k LLM responses to the 62 propositions of the Political Compass Test (PCT) generated by 6 LLMs using 420 prompt variations. We perform coarse-grained analysis of their generated stances and fine-grained analysis of the plain text justifications for those stances. For fine-grained analysis, we propose to identify tropes in the responses: semantically similar phrases that are recurrent and consistent across different prompts, revealing natural patterns in the text that a given LLM is prone to produce. We find that demographic features added to prompts significantly affect outcomes on the PCT, reflecting bias, as well as disparities between the results of tests when eliciting closed-form vs. open domain responses. Additionally, patterns in the plain text rationales via tropes show that similar justifications are repeatedly generated across models and prompts even with disparate stances., Comment: Findings of EMNLP 2024; 28 pages, 20 figures, 7 tables

Chart understanding plays a pivotal role when applying Multimodal Large Language Models (MLLMs) to real-world tasks such as analyzing scientific papers or financial reports. However, existing datasets often focus on oversimplified and homogeneous charts with template-based questions, leading to an over-optimistic measure of progress. We demonstrate that although open-source models can appear to outperform strong proprietary models on these benchmarks, a simple stress test with slightly different charts or questions can deteriorate performance by up to 34.5%. In this work, we propose CharXiv, a comprehensive evaluation suite involving 2,323 natural, challenging, and diverse charts from arXiv papers. CharXiv includes two types of questions: 1) descriptive questions about examining basic chart elements and 2) reasoning questions that require synthesizing information across complex visual elements in the chart. To ensure quality, all charts and questions are handpicked, curated, and verified by human experts. Our results reveal a substantial, previously underestimated gap between the reasoning skills of the strongest proprietary model (i.e., GPT-4o), which achieves 47.1% accuracy, and the strongest open-source model (i.e., InternVL Chat V1.5), which achieves 29.2%. All models lag far behind human performance of 80.5%, underscoring weaknesses in the chart understanding capabilities of existing MLLMs. We hope CharXiv facilitates future research on MLLM chart understanding by providing a more realistic and faithful measure of progress. Project page and leaderboard: https://charxiv.github.io/, Comment: 121 pages, 90 figures

Large language models (LLMs) are used for long-form question answering (LFQA), which requires them to generate paragraph-length answers to complex questions. While LFQA has been well-studied in English, this research has not been extended to other languages. To bridge this gap, we introduce CaLMQA, a collection of 1.5K complex culturally specific questions spanning 23 languages and 51 culturally agnostic questions translated from English into 22 other languages. We define culturally specific questions as those uniquely or more likely to be asked by people from cultures associated with the question's language. We collect naturally-occurring questions from community web forums and hire native speakers to write questions to cover under-resourced, rarely-studied languages such as Fijian and Kirundi. Our dataset contains diverse, complex questions that reflect cultural topics (e.g. traditions, laws, news) and the language usage of native speakers. We automatically evaluate a suite of open- and closed-source models on CaLMQA by detecting incorrect language and token repetitions in answers, and observe that the quality of LLM-generated answers degrades significantly for some low-resource languages. Lastly, we perform human evaluation on a subset of models and languages. Manual evaluation reveals that model performance is significantly worse for culturally specific questions than for culturally agnostic questions. Our findings highlight the need for further research in non-English LFQA and provide an evaluation framework., Comment: 39 pages, 17 figures. Code and data available at https://github.com/2015aroras/CaLMQA. Revised argument in section 4, results unchanged

Decoder-only language models (LMs) have been successfully adopted for speech-processing tasks including automatic speech recognition (ASR). The LMs have ample expressiveness and perform efficiently. This efficiency is a suitable characteristic for streaming applications of ASR. In this work, we propose to use a decoder-only architecture for blockwise streaming ASR. In our approach, speech features are compressed using CTC output and context embedding using blockwise speech subnetwork, and are sequentially provided as prompts to the decoder. The decoder estimates the output tokens promptly at each block. To this end, we also propose a novel training scheme using random-length prefix prompts to make the model robust to the truncated prompts caused by blockwise processing. An experimental comparison shows that our proposed decoder-only streaming ASR achieves 8% relative word error rate reduction in the LibriSpeech test-other set while being twice as fast as the baseline model., Comment: Accepted for Interspeech 2024

Social scientists and the general public often analyze contemporary events by drawing parallels with the past, a process complicated by the vast, noisy, and unstructured nature of historical texts. For example, hundreds of millions of page scans from historical newspapers have been noisily transcribed. Traditional sparse methods for searching for relevant material in these vast corpora, e.g., with keywords, can be brittle given complex vocabularies and OCR noise. This study introduces News Deja Vu, a novel semantic search tool that leverages transformer large language models and a bi-encoder approach to identify historical news articles that are most similar to modern news queries. News Deja Vu first recognizes and masks entities, in order to focus on broader parallels rather than the specific named entities being discussed. Then, a contrastively trained, lightweight bi-encoder retrieves historical articles that are most similar semantically to a modern query, illustrating how phenomena that might seem unique to the present have varied historical precedents. Aimed at social scientists, the user-friendly News Deja Vu package is designed to be accessible for those who lack extensive familiarity with deep learning. It works with large text datasets, and we show how it can be deployed to a massive scale corpus of historical, open-source news articles. While human expertise remains important for drawing deeper insights, News Deja Vu provides a powerful tool for exploring parallels in how people have perceived past and present.

Massive-scale historical document collections are crucial for social science research. Despite increasing digitization, these documents typically lack unique cross-document identifiers for individuals mentioned within the texts, as well as individual identifiers from external knowledgebases like Wikipedia/Wikidata. Existing entity disambiguation methods often fall short in accuracy for historical documents, which are replete with individuals not remembered in contemporary knowledgebases. This study makes three key contributions to improve cross-document coreference resolution and disambiguation in historical texts: a massive-scale training dataset replete with hard negatives - that sources over 190 million entity pairs from Wikipedia contexts and disambiguation pages - high-quality evaluation data from hand-labeled historical newswire articles, and trained models evaluated on this historical benchmark. We contrastively train bi-encoder models for coreferencing and disambiguating individuals in historical texts, achieving accurate, scalable performance that identifies out-of-knowledgebase individuals. Our approach significantly surpasses other entity disambiguation models on our historical newswire benchmark. Our models also demonstrate competitive performance on modern entity disambiguation benchmarks, particularly certain news disambiguation datasets.

Open-Vocabulary Temporal Action Localization (OVTAL) enables a model to recognize any desired action category in videos without the need to explicitly curate training data for all categories. However, this flexibility poses significant challenges, as the model must recognize not only the action categories seen during training but also novel categories specified at inference. Unlike standard temporal action localization, where training and test categories are predetermined, OVTAL requires understanding contextual cues that reveal the semantics of novel categories. To address these challenges, we introduce OVFormer, a novel open-vocabulary framework extending ActionFormer with three key contributions. First, we employ task-specific prompts as input to a large language model to obtain rich class-specific descriptions for action categories. Second, we introduce a cross-attention mechanism to learn the alignment between class representations and frame-level video features, facilitating the multimodal guided features. Third, we propose a two-stage training strategy which includes training with a larger vocabulary dataset and finetuning to downstream data to generalize to novel categories. OVFormer extends existing TAL methods to open-vocabulary settings. Comprehensive evaluations on the THUMOS14 and ActivityNet-1.3 benchmarks demonstrate the effectiveness of our method. Code and pretrained models will be publicly released.

The high luminosity of massive, early-type stars drives strong stellar winds through line scattering of the stars continuum radiation. Their momenta contribute substantially to the dynamics and energetics of the ambient interstellar medium in galaxies. The detailed multi-wavelength study of massive O-type and Wolf-Rayet binaries is essential to explore the hydrodynamics of the shocks formed in the stellar outflows and wind structure. Further, deep analysis of some of the interesting phenomena like particle acceleration and dust formation associated with hot stars winds provides a global view of stellar outflows. In this context, a few massive binaries have been explored using photometric and spectroscopic measurements in different wavebands. This paper highlights important insights gained from investigating massive binaries with several ground and space-based facilities., Comment: 13 pages, 4 figures, Published in the Proceedings of the 3rd BINA Workshop on the Scientific Potential of the Indo-Belgian Cooperation, held in Bhimtal, India

The multiplicity of massive stars is known to be significantly high. Even though the majority of massive stars are located in binary systems, the census of binaries is biased toward shorter periods as longer period systems are more difficult to identify. Alternatively, the search for binary systems with longer periods may proceed differently. As massive binary systems are typically colliding-wind systems, hints for processes occurring in the colliding-wind region could be used as a valuable proxy to identify likely binary systems, and then organize dedicated spectroscopic or interferometric campaigns on a short list of pre-selected targets. In this context, any hint for synchrotron radio emission is seen as a promising indicator of long period binaries, as short period systems undergo severe free-free absorption of the synchrotron emission by the stellar wind material. Usual techniques to identify synchrotron radio emitters constitute thus valid tools to explore that poorly investigated part of the massive binary parameter space. In addition, the identification of a synchrotron emission component in a short period binary can be used as an indicator of the presence of a third companion on a still unrevealed wider orbit in a triple system., Comment: 8 pages, 1 figure, Published in the Proceedings of the 3rd BINA Workshop on the Scientific Potential of the Indo-Belgian Cooperation, held in Bhimtal, India

Recently, multi-task spoken language understanding (SLU) models have emerged, designed to address various speech processing tasks. However, these models often rely on a large number of parameters. Also, they often encounter difficulties in adapting to new data for a specific task without experiencing catastrophic forgetting of previously trained tasks. In this study, we propose finding task-specific subnetworks within a multi-task SLU model via neural network pruning. In addition to model compression, we expect that the forgetting of previously trained tasks can be mitigated by updating only a task-specific subnetwork. We conduct experiments on top of the state-of-the-art multi-task SLU model ``UniverSLU'', trained for several tasks such as emotion recognition (ER), intent classification (IC), and automatic speech recognition (ASR). We show that pruned models were successful in adapting to additional ASR or IC data with minimal performance degradation on previously trained tasks., Comment: Accepted to Interspeech2024

Measurements of the accelerations of stars enabled by time-series extreme-precision spectroscopic observations, from pulsar timing, and from eclipsing binary stars in the Solar Neighborhood offer insights into the mass distribution of the Milky Way that do not rely on traditional equilibrium modeling. Given the measured accelerations, we can determine a total mass density, and from this, by accounting for the mass in stars, gas, and dust, we can infer the amount of dark matter. Leveraging the FIRE-2 simulations of Milky Way-mass galaxies, we compare vertical acceleration profiles between cold dark matter (CDM) and self-interacting dark matter (SIDM) with constant cross-section of 1 cm$^2$ g$^{-1}$ across three halos with diverse assembly histories. Notably, significant asymmetries in vertical acceleration profiles near the midplane at fixed radii are observed in both CDM and SIDM, particularly in halos recently affected by mergers with satellites of Sagittarius/SMC-like masses or greater. These asymmetries offer a unique window into exploring the merger history of a galaxy. We show that SIDM halos consistently exhibit higher local stellar and dark matter densities and steeper vertical acceleration gradients, up to 30% steeper near the Solar Neighborhood. SIDM halos also manifest a more oblate halo shape in the Solar Neighborhood. Furthermore, enhanced precision in acceleration measurements and larger datasets promise to provide better constraints on the local dark matter density, complementing our understanding from kinematic analysis of their distribution within galaxies., Comment: 14 pages, 5 figures, and 2 tables. Submitted to APJ

End-to-end multilingual speech recognition models handle multiple languages through a single model, often incorporating language identification to automatically detect the language of incoming speech. Since the common scenario is where the language is already known, these models can perform as language-specific by using language information as prompts, which is particularly beneficial for attention-based encoder-decoder architectures. However, the Connectionist Temporal Classification (CTC) approach, which enhances recognition via joint decoding and multi-task training, does not normally incorporate language prompts due to its conditionally independent output tokens. To overcome this, we introduce an encoder prompting technique within the self-conditioned CTC framework, enabling language-specific adaptation of the CTC model in a zero-shot manner. Our method has shown to significantly reduce errors by 28% on average and by 41% on low-resource languages., Comment: Accepted by INTERSPEECH 2024

Individuals with diverse motor abilities often benefit from intensive and specialized rehabilitation therapies aimed at enhancing their functional recovery. Nevertheless, the challenge lies in the restricted availability of neurorehabilitation professionals, hindering the effective delivery of the necessary level of care. Robotic devices hold great potential in reducing the dependence on medical personnel during therapy but, at the same time, they generally lack the crucial human interaction and motivation that traditional in-person sessions provide. To bridge this gap, we introduce an AI-based system aimed at delivering personalized, out-of-hospital assistance during neurorehabilitation training. This system includes a rehabilitation training device, affective signal classification models, training exercises, and a socially interactive agent as the user interface. With the assistance of a professional, the envisioned system is designed to be tailored to accommodate the unique rehabilitation requirements of an individual patient. Conceptually, after a preliminary setup and instruction phase, the patient is equipped to continue their rehabilitation regimen autonomously in the comfort of their home, facilitated by a socially interactive agent functioning as a virtual coaching assistant. Our approach involves the integration of an interactive socially-aware virtual agent into a neurorehabilitation robotic framework, with the primary objective of recreating the social aspects inherent to in-person rehabilitation sessions. We also conducted a feasibility study to test the framework with healthy patients. The results of our preliminary investigation indicate that participants demonstrated a propensity to adapt to the system. Notably, the presence of the interactive agent during the proposed exercises did not act as a source of distraction; instead, it positively impacted users' engagement.

The Universal Rotation Curve (URC) of disk galaxies was originally proposed to predict the shape and amplitude of any rotation curve (RC) based solely on photometric data. Here, the URC is investigated with an extensive set of spatially-resolved rotation curves drawn from the PROBES-I, PROBES-II, and MaNGA data bases with matching multi-band surface brightness profiles from the DESI-LIS and WISE surveys for 3,846 disk galaxies. Common URC formulations fail to achieve an adequate level of accuracy to qualify as truly universal over fully sampled RCs. We develop neural network (NN) equivalents for the proposed URCs which predict RCs with higher accuracy, showing that URC inaccuracies are not due to insufficient data but rather non-optimal formulations or sampling effects. This conclusion remains even if the total RC sample is pruned for symmetry. The latest URC prescriptions and their NN equivalents trained on our sub-sample of 579 disk galaxies with symmetric RCs perform similarly to the URC/NN trained on the complete data sample. We conclude that a URC with an acceptable level of accuracy ($\Delta V_{\rm circ} \lesssim15$ per cent) at all radii would require a detailed modelling of a galaxy's central regions and outskirts (e.g., for baryonic effects leading to contraction or expansion of any dark-matter-only halo)., Comment: 36 full pages, 13 figures, 2 appendices. Accepted for publication in the Astrophysical Journal

We introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments with the goal of improving language models. As part of DCLM, we provide a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants in the DCLM benchmark can experiment with data curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline for DCLM, we conduct extensive experiments and find that model-based filtering is key to assembling a high-quality training set. The resulting dataset, DCLM-Baseline enables training a 7B parameter language model from scratch to 64% 5-shot accuracy on MMLU with 2.6T training tokens. Compared to MAP-Neo, the previous state-of-the-art in open-data language models, DCLM-Baseline represents a 6.6 percentage point improvement on MMLU while being trained with 40% less compute. Our baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63% & 66%), and performs similarly on an average of 53 natural language understanding tasks while being trained with 6.6x less compute than Llama 3 8B. Our results highlight the importance of dataset design for training language models and offer a starting point for further research on data curation., Comment: Project page: https://www.datacomp.ai/dclm/

A common method to solve complex problems in software engineering, is to divide the problem into multiple sub-problems. Inspired by this, we propose a Modular Architecture for Software-engineering AI (MASAI) agents, where different LLM-powered sub-agents are instantiated with well-defined objectives and strategies tuned to achieve those objectives. Our modular architecture offers several advantages: (1) employing and tuning different problem-solving strategies across sub-agents, (2) enabling sub-agents to gather information from different sources scattered throughout a repository, and (3) avoiding unnecessarily long trajectories which inflate costs and add extraneous context. MASAI enabled us to achieve the highest performance (28.33% resolution rate) on the popular and highly challenging SWE-bench Lite dataset consisting of 300 GitHub issues from 11 Python repositories. We conduct a comprehensive evaluation of MASAI relative to other agentic methods and analyze the effects of our design decisions and their contribution to the success of MASAI.

The Spoken Language Understanding Evaluation (SLUE) suite of benchmark tasks was recently introduced to address the need for open resources and benchmarking of complex spoken language understanding (SLU) tasks, including both classification and sequence generation tasks, on natural speech. The benchmark has demonstrated preliminary success in using pre-trained speech foundation models (SFM) for these SLU tasks. However, the community still lacks a fine-grained understanding of the comparative utility of different SFMs. Inspired by this, we ask: which SFMs offer the most benefits for these complex SLU tasks, and what is the most effective approach for incorporating these SFMs? To answer this, we perform an extensive evaluation of multiple supervised and self-supervised SFMs using several evaluation protocols: (i) frozen SFMs with a lightweight prediction head, (ii) frozen SFMs with a complex prediction head, and (iii) fine-tuned SFMs with a lightweight prediction head. Although the supervised SFMs are pre-trained on much more speech recognition data (with labels), they do not always outperform self-supervised SFMs; the latter tend to perform at least as well as, and sometimes better than, supervised SFMs, especially on the sequence generation tasks in SLUE. While there is no universally optimal way of incorporating SFMs, the complex prediction head gives the best performance for most tasks, although it increases the inference time. We also introduce an open-source toolkit and performance leaderboard, SLUE-PERB, for these tasks and modeling strategies., Comment: Accepted at ACL Findings 2024

In the U.S. historically, local newspapers drew their content largely from newswires like the Associated Press. Historians argue that newswires played a pivotal role in creating a national identity and shared understanding of the world, but there is no comprehensive archive of the content sent over newswires. We reconstruct such an archive by applying a customized deep learning pipeline to hundreds of terabytes of raw image scans from thousands of local newspapers. The resulting dataset contains 2.7 million unique public domain U.S. newswire articles, written between 1878 and 1977. Locations in these articles are georeferenced, topics are tagged using customized neural topic classification, named entities are recognized, and individuals are disambiguated to Wikipedia using a novel entity disambiguation model. To construct the Newswire dataset, we first recognize newspaper layouts and transcribe around 138 millions structured article texts from raw image scans. We then use a customized neural bi-encoder model to de-duplicate reproduced articles, in the presence of considerable abridgement and noise, quantifying how widely each article was reproduced. A text classifier is used to ensure that we only include newswire articles, which historically are in the public domain. The structured data that accompany the texts provide rich information about the who (disambiguated individuals), what (topics), and where (georeferencing) of the news that millions of Americans read over the course of a century. We also include Library of Congress metadata information about the newspapers that ran the articles on their front pages. The Newswire dataset is useful both for large language modeling - expanding training data beyond what is available from modern web texts - and for studying a diversity of questions in computational linguistics, social science, and the digital humanities., Comment: arXiv admin note: text overlap with arXiv:2306.17810, arXiv:2308.12477

Recent experiments on Cooper pair splitters using superconductor-quantum dot hybrids have embarked on creating entanglement in the solid-state, by engineering the sub-gap processes in the superconducting region. Using the thermoelectric Cooper pair splitter setup [Nat. Comm., 12, 21, (2021)] as a prototype, we develop a detailed analysis of the observed transport signal to bring out vital insights into the regimes of operation and establish the non-local nature of the correlations arising from the crossed Andreev processes. As a striking consequence, contact induced level broadening of the quantum dot's discrete energy spectrum and its hybridization with the superconducting segment, results in a parity reversal of the thermoelectric current along with shifted resonances of the crossed Andreev processes. We conclusively establish the presence of non-local correlations by making a clear nexus with quantum discord. Our detailed analysis thereby provides insights into the gate voltage control of the entanglement generation in superconducting-hybrid Cooper pair splitters., Comment: 10 pages, 4 figures, Comments welcome

The dimer (monomer-dimer) model deals with weighted enumeration of perfect matchings (matchings). The monopole-dimer model is a signed variant of the monomer-dimer model whose partition function is a determinant. In 1999, Lu and Wu evaluated the partition function of the dimer model on two-dimensional grids embedded on a M\"obius strip and a Klein bottle. While the partition function of the dimer model has been known for the two-dimensional grids with different boundary conditions, we present a similar product formula for the partition function of the monopole-dimer model on higher dimensional cylindrical and toroidal grid graphs. We also evaluate the same for the three-dimensional M\"obius and Klein grid graphs and show that the formula does not generalise for the higher dimensions. Further, we present a relation between the product formula for the three-dimensional cylindrical and M\"obius grid., Comment: 25 pages, 9 figures

The current automatic speaker verification (ASV) task involves making binary decisions on two types of trials: target and non-target. However, emerging advancements in speech generation technology pose significant threats to the reliability of ASV systems. This study investigates whether ASV effortlessly acquires robustness against spoofing attacks (i.e., zero-shot capability) by systematically exploring diverse ASV systems and spoofing attacks, ranging from traditional to cutting-edge techniques. Through extensive analyses conducted on eight distinct ASV systems and 29 spoofing attack systems, we demonstrate that the evolution of ASV inherently incorporates defense mechanisms against spoofing attacks. Nevertheless, our findings also underscore that the advancement of spoofing attacks far outpaces that of ASV systems, hence necessitating further research on spoofing-robust ASV methodologies., Comment: 5 pages, 3 figures, 3 tables, Interspeech 2024

Data augmentation has the potential to improve the performance of machine learning models by increasing the amount of training data available. In this study, we evaluated the effectiveness of different data augmentation techniques for a multi-label emotion classification task using a low-resource dataset. Our results showed that Back Translation outperformed autoencoder-based approaches and that generating multiple examples per training instance led to further performance improvement. In addition, we found that Back Translation generated the most diverse set of unigrams and trigrams. These findings demonstrate the utility of Back Translation in enhancing the performance of emotion classification models in resource-limited situations., Comment: The first author contributed significantly

The task of Raga Identification is a very popular research problem in Music Information Retrieval. Few studies that have explored this task employed various approaches, such as signal processing, Machine Learning (ML) methods, and more recently Deep Learning (DL) based methods. However, a key question remains unanswered in all of these works: do these ML/DL methods learn and interpret Ragas in a manner similar to human experts? Besides, a significant roadblock in this research is the unavailability of ample supply of rich, labeled datasets, which drives these ML/DL based methods. In this paper, we introduce "Prasarbharti Indian Music" version-1 (PIM-v1), a novel dataset comprising of 191 hours of meticulously labeled Hindustani Classical Music (HCM) recordings, which is the largest labeled dataset for HCM recordings to the best of our knowledge. Our approach involves conducting ablation studies to find the benchmark classification model for Automatic Raga Identification (ARI) using PIM-v1 dataset. We achieve a chunk-wise f1-score of 0.89 for a subset of 12 Raga classes. Subsequently, we employ model explainability techniques to evaluate the classifier's predictions, aiming to ascertain whether they align with human understanding of Ragas or are driven by arbitrary patterns. We validate the correctness of model's predictions by comparing the explanations given by two ExAI models with human expert annotations. Following this, we analyze explanations for individual test examples to understand the role of regions highlighted by explanations in correct or incorrect predictions made by the model.

In this work, we explore the behavior of interacting dark energy and dark matter within a model of $f(Q)$ gravity, employing a standard framework of dynamical system analysis. We consider the power-law $f(Q)$ model incorporating with two different forms of interacting dark energy and dark matter: $3\alpha H\rho_m$ and $\frac{\alpha}{3H}\rho_m \rho_{DE}$. The evolution of $\Omega_m, \Omega_r, \Omega_{DE}, q$, and $\omega$ for different values of the model parameter $n$ and the interaction parameter $\alpha$ has been examined. Our results show that the universe was dominated by matter in the early stages and will be dominated by dark energy in later stages. Using the observational data, the fixed points are found to be stable and can be represented the de Sitter and quintessence acceleration solutions. We discover that the dynamical profiles of the universe in $f(Q)$ dark energy models are influenced by both the interaction term and the relevant model parameters., Comment: v1: 14 pages, many figures

Let G be a finite group. The nilpotent/commuting/solvable conjugacy class graph is a simple graph with non-central conjugacy classes of G as its vertex set and two vertices are adjacent if and only if a member of one conjugacy class with a member of another conjugacy class generates a nilpotent/abelian/solvable (sub)group. In this paper we have discussed about the forbidden subgraphs of nilpotent, commuting and solvable conjugacy class graphs of groups., Comment: 18 pages

This paper explores the use of 3D lidar in a physical Human-Robot Interaction (pHRI) scenario. To achieve the aforementioned, experiments were conducted to mimic a modern shop-floor environment. Data was collected from a pool of seventeen participants while performing pre-determined tasks in a shared workspace with the robot. To demonstrate an end-to-end case; a perception pipeline was developed that leverages reflectivity, signal, near-infrared, and point-cloud data from a 3-D lidar. This data is then used to perform safety based control whilst satisfying the speed and separation monitoring (SSM) criteria. In order to support the perception pipeline, a state-of-the-art object detection network was leveraged and fine-tuned by transfer learning. An analysis is provided along with results of the perception and the safety based controller. Additionally, this system is compared with the previous work., Comment: Submitted to IEEE-CASE 2024. Under Review

This paper discusses the results of the third edition of the Monocular Depth Estimation Challenge (MDEC). The challenge focuses on zero-shot generalization to the challenging SYNS-Patches dataset, featuring complex scenes in natural and indoor settings. As with the previous edition, methods can use any form of supervision, i.e. supervised or self-supervised. The challenge received a total of 19 submissions outperforming the baseline on the test set: 10 among them submitted a report describing their approach, highlighting a diffused use of foundational models such as Depth Anything at the core of their method. The challenge winners drastically improved 3D F-Score performance, from 17.51% to 23.72%., Comment: To appear in CVPRW2024