Search

Low-disorder two-dimensional electron systems in the presence of a strong, perpendicular magnetic field terminate at very small Landau level filling factors in a Wigner crystal (WC), where the electrons form an ordered array to minimize the Coulomb repulsion. The nature of this exotic, many-body, quantum phase is yet to be fully understood and experimentally revealed. Here we probe one of WC's most fundamental parameters, namely the energy gap that determines its low-temperature conductivity, in record-mobility, ultra-high-purity, two-dimensional electrons confined to GaAs quantum wells. The WC domains in these samples contain $\simeq$ 1000 electrons. The measured gaps are a factor of three larger than previously reported for lower quality samples, and agree remarkably well with values predicted for the lowest-energy, intrinsic, hyper-corelated bubble defects in a WC made of flux-electron composite fermions, rather than bare electrons. The agreement is particularly noteworthy, given that the calculations are done for disorder-free composite fermion WCs, and there are no adjustable parameters. The results reflect the exceptionally high quality of the samples, and suggest that composite fermion WCs are indeed more stable compared to their electron counterparts.

Low dimensional hybrid organic-inorganic perovskites (HOIPs) represent a promising class of electronically active materials for both light absorption and emission. The design space of HOIPs is extremely large, since a diverse space of organic cations can be combined with different inorganic frameworks. This immense design space allows for tunable electronic and mechanical properties, but also necessitates the development of new tools for in silico high throughput analysis of candidate structures. In this work, we present an accurate, efficient, transferable and widely applicable machine learning interatomic potential (MLIP) for predicting the structure of new 2D HOIPs. Using the MACE architecture, an MLIP is trained on 86 diverse experimentally reported HOIP structures. The model is tested on 73 unseen perovskite compositions, and achieves chemical accuracy with respect to the reference electronic structure method. Our model is then combined with a simple random structure search algorithm to predict the structure of hypothetical HOIPs given only the proposed composition. Success is demonstrated by correctly and reliably recovering the crystal structure of a set of experimentally known 2D perovskites. Such a random structure search is impossible with ab initio methods due to the associated computational cost, but is relatively inexpensive with the MACE potential. Finally, the procedure is used to predict the structure formed by a new organic cation with no previously known corresponding perovskite. Laboratory synthesis of the new hybrid perovskite confirms the accuracy of our prediction. This capability, applied at scale, enables efficient screening of thousands of combinations of organic cations and inorganic layers., Comment: 14 pages and 9 figures in the main text. Supplementary included in pdf

Large language models (LLMs) are notorious for hallucinating, i.e., producing erroneous claims in their output. Such hallucinations can be dangerous, as occasional factual inaccuracies in the generated text might be obscured by the rest of the output being generally factually correct, making it extremely hard for the users to spot them. Current services that leverage LLMs usually do not provide any means for detecting unreliable generations. Here, we aim to bridge this gap. In particular, we propose a novel fact-checking and hallucination detection pipeline based on token-level uncertainty quantification. Uncertainty scores leverage information encapsulated in the output of a neural network or its layers to detect unreliable predictions, and we show that they can be used to fact-check the atomic claims in the LLM output. Moreover, we present a novel token-level uncertainty quantification method that removes the impact of uncertainty about what claim to generate on the current step and what surface form to use. Our method Claim Conditioned Probability (CCP) measures only the uncertainty of a particular claim value expressed by the model. Experiments on the task of biography generation demonstrate strong improvements for CCP compared to the baselines for seven LLMs and four languages. Human evaluation reveals that the fact-checking pipeline based on uncertainty quantification is competitive with a fact-checking tool that leverages external knowledge., Comment: Accepted to ACL-2024 (Findings). Ekaterina Fadeeva, Aleksandr Rubashevskii, and Artem Shelmanov contributed equally

AB oscillation in weak magnetic field (B<1.5kG) is observed in QPC due to interference between electrons propagating along different QPC channels. We also investigate photo-induced magnetoresistance oscillation in open-regime split-gate QPC under MW irradiation. It is attributed to EMPs interfering in the QPC region. The influence of MW power, frequency and split gate voltage is discussed thoroughly. We unify the result of photoconductance at B=0 with EMP theories.

We measure the resistance oscillation of quantum point contact (QPC) under microwave (MW) radiation. What is different from the common resistance oscillation induced by edge magnetoplasmon (EMP) is that at lower magnetic field ({\omega}>{\omega}_c), photoconductance is positive (negative) with weak (strong) MW radiation. This transport phenomenon is proved to be related to superballistic flow of electrons through QPC. The distinction between the regions {\omega}>{\omega}_c and {\omega}<{\omega}_c is attributed to different absorption rate of MW radiation. Violent absorption occurs when cyclotron orbits or current domains are destroyed in QPC region.

We systematically study the acousto-current of two-dimensional electron systems in the integer and fractional quantum Hall regimes using surface acoustic waves. We are able to separate the co-existing acoustic scattering and drag, when phonons induce drag current and tune the electron conductivity, respectively. At large acoustic power, the drag current is finite when the system is compressible and exhibits minima when incompressible quantum Hall effects appear. Surprisingly, it exhibits anomalously large bipolar spikes within the quantum Hall plateaus while it vanishes linearly with reduced acoustic power at compressible phases. The current peaks reverse their polarity at the two flanks of exact integer or fractional fillings, consistent with the opposite electric charge of the quasiparticle/quasihole.

"Bigger the better" has been the predominant trend in recent Large Language Models (LLMs) development. However, LLMs do not suit well for scenarios that require on-device processing, energy efficiency, low memory footprint, and response efficiency. These requisites are crucial for privacy, security, and sustainable deployment. This paper explores the "less is more" paradigm by addressing the challenge of designing accurate yet efficient Small Language Models (SLMs) for resource constrained devices. Our primary contribution is the introduction of an accurate and fully transparent open-source 0.5 billion (0.5B) parameter SLM, named MobiLlama, catering to the specific needs of resource-constrained computing with an emphasis on enhanced performance with reduced resource demands. MobiLlama is a SLM design that initiates from a larger model and applies a careful parameter sharing scheme to reduce both the pre-training and the deployment cost. Our work strives to not only bridge the gap in open-source SLMs but also ensures full transparency, where complete training data pipeline, training code, model weights, and over 300 checkpoints along with evaluation codes is available at : https://github.com/mbzuai-oryx/MobiLlama., Comment: Code available at : https://github.com/mbzuai-oryx/MobiLlama

In pursuit of more inclusive Vision-Language Models (VLMs), this study introduces a Large Multilingual Multimodal Model called PALO. PALO offers visual reasoning capabilities in 10 major languages, including English, Chinese, Hindi, Spanish, French, Arabic, Bengali, Russian, Urdu, and Japanese, that span a total of ~5B people (65% of the world population). Our approach involves a semi-automated translation approach to adapt the multimodal instruction dataset from English to the target languages using a fine-tuned Large Language Model, thereby ensuring high linguistic fidelity while allowing scalability due to minimal manual effort. The incorporation of diverse instruction sets helps us boost overall performance across multiple languages especially those that are underrepresented like Hindi, Arabic, Bengali, and Urdu. The resulting models are trained across three scales (1.7B, 7B and 13B parameters) to show the generalization and scalability where we observe substantial improvements compared to strong baselines. We also propose the first multilingual multimodal benchmark for the forthcoming approaches to evaluate their vision-language reasoning capabilities across languages. Code: https://github.com/mbzuai-oryx/PALO., Comment: Technical Report of PALO

Recent studies have demonstrated that large language models (LLMs) have ethical-related problems such as social biases, lack of moral reasoning, and generation of offensive content. The existing evaluation metrics and methods to address these ethical challenges use datasets intentionally created by instructing humans to create instances including ethical problems. Therefore, the data does not reflect prompts that users actually provide when utilizing LLM services in everyday contexts. This may not lead to the development of safe LLMs that can address ethical challenges arising in real-world applications. In this paper, we create Eagle datasets extracted from real interactions between ChatGPT and users that exhibit social biases, toxicity, and immoral problems. Our experiments show that Eagle captures complementary aspects, not covered by existing datasets proposed for evaluation and mitigation of such ethical challenges. Our code is publicly available at https://huggingface.co/datasets/MasahiroKaneko/eagle.

The focus of language model evaluation has transitioned towards reasoning and knowledge-intensive tasks, driven by advancements in pretraining large models. While state-of-the-art models are partially trained on large Arabic texts, evaluating their performance in Arabic remains challenging due to the limited availability of relevant datasets. To bridge this gap, we present \datasetname{}, the first multi-task language understanding benchmark for the Arabic language, sourced from school exams across diverse educational levels in different countries spanning North Africa, the Levant, and the Gulf regions. Our data comprises 40 tasks and 14,575 multiple-choice questions in Modern Standard Arabic (MSA) and is carefully constructed by collaborating with native speakers in the region. Our comprehensive evaluations of 35 models reveal substantial room for improvement, particularly among the best open-source models. Notably, BLOOMZ, mT0, LLaMA2, and Falcon struggle to achieve a score of 50%, while even the top-performing Arabic-centric model only achieves a score of 62.3%., Comment: Findings of ACL 2024

In this paper, we introduce BiMediX, the first bilingual medical mixture of experts LLM designed for seamless interaction in both English and Arabic. Our model facilitates a wide range of medical interactions in English and Arabic, including multi-turn chats to inquire about additional details such as patient symptoms and medical history, multiple-choice question answering, and open-ended question answering. We propose a semi-automated English-to-Arabic translation pipeline with human refinement to ensure high-quality translations. We also introduce a comprehensive evaluation benchmark for Arabic medical LLMs. Furthermore, we introduce BiMed1.3M, an extensive Arabic-English bilingual instruction set covering 1.3 Million diverse medical interactions, resulting in over 632 million healthcare specialized tokens for instruction tuning. Our BiMed1.3M dataset includes 250k synthesized multi-turn doctor-patient chats and maintains a 1:2 Arabic-to-English ratio. Our model outperforms state-of-the-art Med42 and Meditron by average absolute gains of 2.5% and 4.1%, respectively, computed across multiple medical evaluation benchmarks in English, while operating at 8-times faster inference. Moreover, our BiMediX outperforms the generic Arabic-English bilingual LLM, Jais-30B, by average absolute gains of 10% on our Arabic medical benchmark and 15% on bilingual evaluations across multiple datasets. Our project page with source code and trained model is available at https://github.com/mbzuai-oryx/BiMediX ., Comment: Accepted to EMNLP 2024 (Findings)

The world's languages exhibit certain so-called typological or implicational universals; for example, Subject-Object-Verb (SOV) languages typically use postpositions. Explaining the source of such biases is a key goal of linguistics. We study word-order universals through a computational simulation with language models (LMs). Our experiments show that typologically-typical word orders tend to have lower perplexity estimated by LMs with cognitively plausible biases: syntactic biases, specific parsing strategies, and memory limitations. This suggests that the interplay of cognitive biases and predictability (perplexity) can explain many aspects of word-order universals. It also showcases the advantage of cognitively-motivated LMs, typically employed in cognitive modeling, in the simulation of language universals., Comment: Accepted by ACL 2024 main conference, 22 pages

Many studies have demonstrated that large language models (LLMs) can produce harmful responses, exposing users to unexpected risks when LLMs are deployed. Previous studies have proposed comprehensive taxonomies of the risks posed by LLMs, as well as corresponding prompts that can be used to examine the safety mechanisms of LLMs. However, the focus has been almost exclusively on English, and little has been explored for other languages. Here we aim to bridge this gap. We first introduce a dataset for the safety evaluation of Chinese LLMs, and then extend it to two other scenarios that can be used to better identify false negative and false positive examples in terms of risky prompt rejections. We further present a set of fine-grained safety assessment criteria for each risk type, facilitating both manual annotation and automatic evaluation in terms of LLM response harmfulness. Our experiments on five LLMs show that region-specific risks are the prevalent type of risk, presenting the major issue with all Chinese LLMs we experimented with. Our data is available at https://github.com/Libr-AI/do-not-answer. Warning: this paper contains example data that may be offensive, harmful, or biased., Comment: 14 pages

Large language models (LLMs) have achieved success in acting as agents, which interact with environments through tools such as search engines. However, LLMs are optimized for language generation instead of tool use during training or alignment, limiting their effectiveness as agents. To resolve this problem, previous work has first collected interaction trajectories between LLMs and environments, using only trajectories that successfully finished the task to fine-tune smaller models, making fine-tuning data scarce and acquiring it both difficult and costly. Discarding failed trajectories also leads to significant wastage of data and resources and limits the possible optimization paths during fine-tuning. In this paper, we argue that unsuccessful trajectories offer valuable insights, and LLMs can learn from these trajectories through appropriate quality control and fine-tuning strategies. By simply adding a prefix or suffix that tells the model whether to generate a successful trajectory during training, we improve model performance by a large margin on mathematical reasoning, multi-hop question answering, and strategic question answering tasks. We further analyze the inference results and find that our method provides a better trade-off between valuable information and errors in unsuccessful trajectories. To our knowledge, we are the first to demonstrate the value of negative trajectories and their application in agent-tunning scenarios. Our findings offer guidance for developing better agent-tuning methods and low-resource data usage techniques., Comment: Agent, LLM, Large Language Model

The composite fermion theory opened a new chapter in understanding many-body correlations through the formation of emergent particles. The formation of two-flux and four-flux composite fermions is well established. While there are limited data linked to the formation of six-flux composite fermions, topological protection associated with them is conspicuously lacking. Here we report evidence for the formation of a quantized and gapped fractional quantum Hall state at the filling factor $\nu=9/11$, which we associate with the formation of six-flux composite fermions. Our result provides evidence for the most intricate composite fermion with six fluxes and expands the already diverse family of highly correlated topological phases with a new member that cannot be characterized by correlations present in other known members. Our observations pave the way towards the study of higher order correlations in the fractional quantum Hall regime.

Traditional language models, adept at next-token prediction in text sequences, often struggle with transduction tasks between distinct symbolic systems, particularly when parallel data is scarce. Addressing this issue, we introduce \textit{symbolic autoencoding} ($\Sigma$AE), a self-supervised framework that harnesses the power of abundant unparallel data alongside limited parallel data. $\Sigma$AE connects two generative models via a discrete bottleneck layer and is optimized end-to-end by minimizing reconstruction loss (simultaneously with supervised loss for the parallel data), such that the sequence generated by the discrete bottleneck can be read out as the transduced input sequence. We also develop gradient-based methods allowing for efficient self-supervised sequence learning despite the discreteness of the bottleneck. Our results demonstrate that $\Sigma$AE significantly enhances performance on transduction tasks, even with minimal parallel data, offering a promising solution for weakly supervised learning scenarios.

This paper describes the energy resolution of Microwave Kinetic Inductance Detectors (MKIDs), and models some limiting factors to it. Energy resolution is a measure of the smallest possible difference in energy of the impinging photons, Delta E, that the detector can identify and is of critical importance for many applications. Limits to the energy resolution cause by the Fano effect, amplifier noise, current inhomogeneities, and readout sampling frequency are taken into consideration for this model. This paper describes an approach to combine all of these limitations and predict a wavelength dependency of the upper limit to the resolving power.

The QED-C suite of Application-Oriented Benchmarks provides the ability to gauge performance characteristics of quantum computers as applied to real-world applications. Its benchmark programs sweep over a range of problem sizes and inputs, capturing key performance metrics related to the quality of results, total time of execution, and quantum gate resources consumed. In this manuscript, we investigate challenges in broadening the relevance of this benchmarking methodology to applications of greater complexity. First, we introduce a method for improving landscape coverage by varying algorithm parameters systematically, exemplifying this functionality in a new scalable HHL linear equation solver benchmark. Second, we add a VQE implementation of a Hydrogen Lattice simulation to the QED-C suite, and introduce a methodology for analyzing the result quality and run-time cost trade-off. We observe a decrease in accuracy with increased number of qubits, but only a mild increase in the execution time. Third, unique characteristics of a supervised machine-learning classification application are explored as a benchmark to gauge the extensibility of the framework to new classes of application. Applying this to a binary classification problem revealed the increase in training time required for larger anzatz circuits, and the significant classical overhead. Fourth, we add methods to include optimization and error mitigation in the benchmarking workflow which allows us to: identify a favourable trade off between approximate gate synthesis and gate noise; observe the benefits of measurement error mitigation and a form of deterministic error mitigation algorithm; and to contrast the improvement with the resulting time overhead. Looking ahead, we discuss how the benchmark framework can be instrumental in facilitating the exploration of algorithmic options and their impact on performance., Comment: 21 pages, 21 figures

Improving multilingual language models capabilities in low-resource languages is generally difficult due to the scarcity of large-scale data in those languages. In this paper, we relax the reliance on texts in low-resource languages by using multilingual lexicons in pretraining to enhance multilingual capabilities. Specifically, we focus on zero-shot sentiment analysis tasks across 34 languages, including 6 high/medium-resource languages, 25 low-resource languages, and 3 code-switching datasets. We demonstrate that pretraining using multilingual lexicons, without using any sentence-level sentiment data, achieves superior zero-shot performance compared to models fine-tuned on English sentiment datasets, and large language models like GPT--3.5, BLOOMZ, and XGLM. These findings are observable for unseen low-resource languages to code-mixed scenarios involving high-resource languages., Comment: Accepted at EACL 2024

There exist both scalable tasks, like reading comprehension and fact-checking, where model performance improves with model size, and unscalable tasks, like arithmetic reasoning and symbolic reasoning, where model performance does not necessarily improve with model size. Large language models (LLMs) equipped with Chain-of-Thought (CoT) prompting are able to make accurate incremental predictions even on unscalable tasks. Unfortunately, despite their exceptional reasoning abilities, LLMs tend to internalize and reproduce discriminatory societal biases. Whether CoT can provide discriminatory or egalitarian rationalizations for the implicit information in unscalable tasks remains an open question. In this study, we examine the impact of LLMs' step-by-step predictions on gender bias in unscalable tasks. For this purpose, we construct a benchmark for an unscalable task where the LLM is given a list of words comprising feminine, masculine, and gendered occupational words, and is required to count the number of feminine and masculine words. In our CoT prompts, we require the LLM to explicitly indicate whether each word in the word list is a feminine or masculine before making the final predictions. With counting and handling the meaning of words, this benchmark has characteristics of both arithmetic reasoning and symbolic reasoning. Experimental results in English show that without step-by-step prediction, most LLMs make socially biased predictions, despite the task being as simple as counting words. Interestingly, CoT prompting reduces this unconscious social bias in LLMs and encourages fair predictions.

In low-disorder, two-dimensional electron systems (2DESs), the fractional quantum Hall states at very small Landau level fillings ($\nu$) terminate in a Wigner solid (WS) phase, where electrons arrange themselves in a periodic array. The WS is typically pinned by the residual disorder sites and manifests an insulating behavior, with non-linear current-voltage (\textit{I-V}) and noise characteristics. We report here, measurements on an ultra-low-disorder, dilute 2DES, confined to a GaAs quantum well. In the $\nu < 1/5$ range, superimposed on a highly-insulating longitudinal resistance, the 2DES exhibits a developing fractional quantum Hall state at $\nu=1/7$, attesting to its exceptional high quality, and dominance of electron-electron interaction in the low filling regime. In the nearby insulating phases, we observe remarkable non-linear \textit{I-V} and noise characteristics as a function of increasing current, with current thresholds delineating three distinct phases of the WS: a pinned phase (P1) with very small noise, a second phase (P2) in which $dV/dI$ fluctuates between positive and negative values and is accompanied by very high noise, and a third phase (P3) where $dV/dI$ is nearly constant and small, and noise is about an order of magnitude lower than in P2. In the depinned (P2 and P3) phases, the noise spectrum also reveals well-defined peaks at frequencies that vary linearly with the applied current, suggestive of washboard frequencies. We discuss the data in light of a recent theory that proposes different dynamic phases for a driven WS.

We study dipolar excitons confined at 330 mK in a square electrostatic lattice of a GaAs double quantum well. In the dipolar occupation blockade regime, at 3/2 filling, we evidence that excitons form a face-centred superlattice quantum solid. This phase is realised with high purity across 36 lattice sites, in a regime where the excitons mean interaction energy exceeds the depth of the lattice confinement. The superlattice solid then closely relates to Wigner crystals., Comment: 10 pages, 5 figures

GN-z11 is the highest redshift galaxy spectroscopically confirmed with the Hubble Space Telescope (HST). Previous measurements of the effective radius of GN-z11 utilized galfit, which is not optimized to measure structural parameters for such a faint, distant object. Using a new software program called forcepho on HST data for the first time, we derive a size from images in the F160W band obtained both from the complete CANDELS survey and additional midcycle observations in order to contribute to the knowledge base on the size evolution, size-luminosity, and size-mass relation of early galaxies. We find a half-light radius mean of 0''.036 \(\pm\) 0''.006 corresponding to a physical size of 0.15 \(\pm\) 0.025 kpc. This size, smaller than the point spread function, is dramatically smaller than previous estimates with shallower HST data using galfit but consistent with recent measurements using forcepho on new JWST data arXiv:2302.07234. Such a small size, combined with the JWST/NIRSpec spectroscopic observations arXiv:2305.12492, suggests that GN-z11's high luminosity is dominated by an AGN., Comment: 4 pages, 1 figure

Answering real-world tourism questions that seek Point-of-Interest (POI) recommendations is challenging, as it requires both spatial and non-spatial reasoning, over a large candidate pool. The traditional method of encoding each pair of question and POI becomes inefficient when the number of candidates increases, making it infeasible for real-world applications. To overcome this, we propose treating the QA task as a dense vector retrieval problem, where we encode questions and POIs separately and retrieve the most relevant POIs for a question by utilizing embedding space similarity. We use pretrained language models (PLMs) to encode textual information, and train a location encoder to capture spatial information of POIs. Experiments on a real-world tourism QA dataset demonstrate that our approach is effective, efficient, and outperforms previous methods across all metrics. Enabled by the dense retrieval architecture, we further build a global evaluation baseline, expanding the search space by 20 times compared to previous work. We also explore several factors that impact on the model's performance through follow-up experiments. Our code and model are publicly available at https://github.com/haonan-li/LAMB., Comment: Accepted at AACL 2023

Machine-learned force fields have transformed the atomistic modelling of materials by enabling simulations of ab initio quality on unprecedented time and length scales. However, they are currently limited by: (i) the significant computational and human effort that must go into development and validation of potentials for each particular system of interest; and (ii) a general lack of transferability from one chemical system to the next. Here, using the state-of-the-art MACE architecture we introduce a single general-purpose ML model, trained on a public database of 150k inorganic crystals, that is capable of running stable molecular dynamics on molecules and materials. We demonstrate the power of the MACE-MP-0 model - and its qualitative and at times quantitative accuracy - on a diverse set problems in the physical sciences, including the properties of solids, liquids, gases, chemical reactions, interfaces and even the dynamics of a small protein. The model can be applied out of the box and as a starting or "foundation model" for any atomistic system of interest and is thus a step towards democratising the revolution of ML force fields by lowering the barriers to entry., Comment: 119 pages, 63 figures, 37MB PDF

We present very-long-baseline interferometry (VLBI) observations of the hyperactive repeating FRB 20220912A using the European VLBI Network (EVN) with an EVN-Lite setup. We detected 150 bursts from FRB 20220912A over two observing epochs in October 2022. Combining the data of these bursts allows us to localise FRB 20220912A to a precision of a few milliarcseconds, corresponding to a transverse scale of less than 10 pc at the distance of the source. The precision of this localisation shows that FRB 20220912A lies closer to the centre of its host galaxy than previously found, although still significantly offset from the host galaxy's nucleus. On arcsecond scales, FRB 20220912A is coincident with a persistent continuum radio source known from archival observations, however, we find no compact persistent emission on milliarcsecond scales. The persistent radio emission is thus likely to be from star-formation in the host galaxy. This is in contrast to some other active FRBs, such as FRB 20121102A and FRB 20190520B., Comment: 13 pages, 7 figures, submitted to MNRAS, comments most welcome

Owing to their large effective mass, strong and tunable spin-orbit coupling, and complex band-structure, two-dimensional hole systems (2DHSs) in GaAs quantum wells provide rich platforms to probe exotic many-body physics, while also offering potential applications in ballistic and spintronics devices, and fault-tolerant topological quantum computing. We present here a systematic study of molecular-beam-epitaxy grown, modulation-doped, GaAs (001) 2DHSs where we explore the limits of low-temperature 2DHS mobility by optimizing two parameters, the GaAs quantum well width and the alloy fraction ($x$) of the flanking Al$_x$Ga$_{1-x}$As barriers. We obtain a breakthrough in 2DHS mobility, with a peak value $\simeq 18 \times 10^6$ cm$^2$/Vs at a density of 3.8 $\times$ 10$^{10}$ /cm$^{2}$, implying a mean-free-path of $\simeq 57 \mu$m. Using transport calculations tailored to our structures, we analyze the operating scattering mechanisms to explain the non-monotonic evolution of mobility with density. We find it imperative to include the dependence of effective mass on 2DHS density, well width, and $x$. We observe concomitant improvement in quality as evinced by the appearance of delicate fractional quantum Hall states at very low density., Comment: 10 pages, 4 figures, Accepted in Physical Review Materials

Inspired by considerations in $M$-theory, we prove the equivalence between the moduli spaces of (suitably complexified) torsion free $G_{2}$-structures on 7-manifolds which are families of hyperK\"ahler ALE 4-manifolds fibered over compact flat 3-manifolds $T^{3}/K$ and certain moduli spaces of flat $G^{\mathbb{C}}_{ADE}$ connections on $T^{3}/K$.

Instruction tuning significantly enhances the performance of large language models (LLMs) across various tasks. However, the procedure to optimizing the mixing of instruction datasets for LLM fine-tuning is still poorly understood. This study categorizes instructions into three primary types: NLP downstream tasks, coding, and general chat. We explore the effects of instruction tuning on different combinations of datasets on LLM performance, and find that certain instruction types are more advantageous for specific applications but can negatively impact other areas. This work provides insights into instruction mixtures, laying the foundations for future research., Comment: Instruction Tuning, Large Language Model, Alignment

We present new high-sensitivity e-MERLIN and VLA radio images of the prototypical Seyfert 2 galaxy NGC 1068 at 5, 10 and 21 GHz. We image the radio jet, from the compact components NE, C, S1 and S2 to the faint double-lobed jet structure of the NE and SW jet lobes. Furthermore, we map the jet between by combining e-MERLIN and VLA data for the first time. Components NE, C and S2 have steep spectra indicative of optically-thin non-thermal emission domination between 5 and 21 GHz. Component S1, which is where the AGN resides, has a flat radio spectrum. We report a new component, S2a, a part of the southern jet. We compare these new data with the MERLIN and VLA data observed in 1983, 1992 and 1995 and report a flux decrease by a factor of 2 in component C, suggesting variability of this jet component. With the high angular resolution e-MERLIN maps, we detect the bow shocks in the NE jet lobe that coincide with the molecular gas outflows observed with ALMA. The NE jet lobe has enough radio power considered to be responsible for driving out the dense molecular gas observed with ALMA around the same region.

The recent surge in open-source Large Language Models (LLMs), such as LLaMA, Falcon, and Mistral, provides diverse options for AI practitioners and researchers. However, most LLMs have only released partial artifacts, such as the final model weights or inference code, and technical reports increasingly limit their scope to high-level design choices and surface statistics. These choices hinder progress in the field by degrading transparency into the training of LLMs and forcing teams to rediscover many details in the training process. We present LLM360, an initiative to fully open-source LLMs, which advocates for all training code and data, model checkpoints, and intermediate results to be made available to the community. The goal of LLM360 is to support open and collaborative AI research by making the end-to-end LLM training process transparent and reproducible by everyone. As a first step of LLM360, we release two 7B parameter LLMs pre-trained from scratch, Amber and CrystalCoder, including their training code, data, intermediate checkpoints, and analyses (at https://www.llm360.ai). We are committed to continually pushing the boundaries of LLMs through this open-source effort. More large-scale and stronger models are underway and will be released in the future.