Search

Instruction tuning is a supervised fine-tuning approach that significantly improves the ability of large language models (LLMs) to follow human instructions. We propose SelfCodeAlign, the first fully transparent and permissive pipeline for self-aligning code LLMs without extensive human annotations or distillation. SelfCodeAlign employs the same base model for inference throughout the data generation process. It first extracts diverse coding concepts from high-quality seed snippets to generate new tasks. It then samples multiple responses per task, pairs each with test cases, and validates them in a sandbox environment. Finally, passing examples are selected for instruction tuning. In our primary experiments, we use SelfCodeAlign with CodeQwen1.5-7B to generate a dataset of 74k instruction-response pairs. Finetuning on this dataset leads to a model that achieves a 67.1 pass@1 on HumanEval+, surpassing CodeLlama-70B-Instruct despite being ten times smaller. Across all benchmarks, this finetuned model consistently outperforms the original version trained with OctoPack, the previous state-of-the-art method for instruction tuning without human annotations or distillation. Additionally, we show that SelfCodeAlign is effective across LLMs of various sizes, from 3B to 33B, and that the base models can benefit more from alignment with their own data distribution. We further validate each component's effectiveness in our pipeline, showing that SelfCodeAlign outperforms both direct distillation from GPT-4o and leading GPT-3.5-based distillation methods, such as OSS-Instruct and Evol-Instruct. SelfCodeAlign has also led to the creation of StarCoder2-Instruct, the first fully transparent, permissively licensed, and self-aligned code LLM that achieves state-of-the-art coding performance., Comment: Accepted to NeurIPS 2024

The Perseus cluster is the brightest X-ray cluster in the sky and is known as a cool-core galaxy cluster. Being a very nearby cluster, it has been extensively studied. This has provided a comprehensive view of the physical processes that operate in the intracluster medium (ICM), including feedback from the AGN 3C84 and measurements of ICM turbulence. Additionally, the Perseus cluster contains a central radio mini-halo. This diffuse radio source traces cosmic ray electrons (re-)accelerated in-situ in the ICM. Here we report on LOFAR high-band antenna 120-168 MHz observations of the Perseus cluster that probe a range of four orders of magnitude in angular scales. In our 0.3 arcsec resolution image, we find that the northern extension of the 3C84 lobe consists of several narrow 1.5-3 kpc parallel strands of emission. In addition, we detect steep-spectrum filaments associated with a previous outburst of the central AGN radio emission filling two known X-ray ghost cavities. At 7 arcsec resolution, our images show a complex structured radio mini-halo, with several edges and filaments. At resolutions of 26 arcsec and 80 arcsec, we discover diffuse radio emission with a 1.1 Mpc extent. We classify this emission as a giant radio halo and its properties are distinct from the inner mini-halo. We also detect two diffuse sources at projected cluster centric radii of 0.7 and 1.0 Mpc. Finally, we observe a 0.9 Mpc long trail of radio emission from the cluster member galaxy IC310, connecting it with the giant radio halo. Together with other recent studies of relaxed clusters, our LOFAR observations indicate that cluster-wide radio emission could be (more) common in cool-core clusters. In the case of the Perseus cluster, a past off-axis merger event that preserved the cool core might have generated enough turbulence to produce an extended radio halo observable at low frequencies., Comment: accepted for publication in A&A, 18 pages, 19 figures

We present the first multi-frequency analysis of the candidate ultra-steep spectrum radio halo in the galaxy cluster PLCKESZ G171.94$-$40.65, using the upgraded Giant Metrewave Radio telescope (uGMRT; 400 MHz), and Karl G. Jansky Very Large Array (JVLA; 1-2 GHz) observations. Our radio data have been complemented with archival \textit{Chandra} X-ray observations to provide a crucial insight into the complex intracluster medium (ICM) physics, happening at large scales. We detect the radio halo emission to the extent of $\sim$ 1.5 Mpc at 400 MHz, significantly larger than previously reported, along with five tailed galaxies in the central region. We also report the discovery of an unknown diffuse source 'U', at the cluster periphery, with an extent of 300 kpc. Using the available observations, we have found that the radio spectrum of the halo is well-fitted with a single power law, having a spectral index of $-1.36 \pm 0.05$, indicating that it is not an ultra-steep spectrum radio halo. Our low-resolution (25$''$) resolved spectral map shows an overall uniform spectral index, with some patches of fluctuations. The X-ray and radio surface brightness are morphologically co-spatial, with a slight extension along the northwest-southeast direction, seen in both maps. The radio and X-ray surface brightness indicates strong positive correlations, with sub-linear correlation slopes ($\sim$ 0.71). Multiple tailed galaxies and the radio halo indicate a high dynamical activity at the cluster central region., Comment: 16 pages, 12 Figures, 6 Tables, Accepted for publication in ApJ

The clusters Abell 2061 and Abell 2067 in the Corona Borealis supercluster have been studied at different radio frequencies and are both known to host diffuse radio emission. The aim of this work is to investigate the radio emission in between them, suggested by low resolution observations. We analyse deep LOFAR HBA observations at 144 MHz to follow up on the possible intercluster filament suggested by previous 1.4 GHz observations. We investigate the radial profiles and the point-to-point surface brightness correlation of the emission in A2061 with radio and Xray observations, to describe the nature of the diffuse emission. We report the detection of diffuse radio emission on 800 kpc scale, more extended than previously known, reaching beyond the radio halo in A2061 towards A2067 and over the separation outside the two clusters R500 radii. We confirm the presence of a radio halo in A2061, while do not find evidence of diffuse emission in A2067. The surface brightness profile from the centre of A2061 shows an excess of emission with respect to the azimuthally averaged radio halo profile and X-ray background. We explore three different dynamical scenario to explain the nature of the diffuse emission. We analyse a trail of emission of 760 kpc between the radio halo and radio relic in A2061. This pre merger system closely resembles the two other cluster pairs where radio bridges connecting the radio halos on Mpc scales have been detected. The diffuse emission extends beyond each cluster R500 radius but in this unique case, the absence of the radio halo in A2067 is likely the reason for the observed 'gap' between the two systems. However, the point-to-point correlation results are challenging to explain. The classification of the emission remains unclear, and detailed spectral analysis and further Xray observations are required to understand the origin of the diffuse emission., Comment: Accepted for publication on Astronomy & Astrophysics 12 pages, 6 figures, 1 appendix

While scaling training compute has led to remarkable improvements in large language models (LLMs), scaling inference compute has not yet yielded analogous gains. We hypothesize that a core missing component is a lack of diverse LLM outputs, leading to inefficient search due to models repeatedly sampling highly similar, yet incorrect generations. We empirically demonstrate that this lack of diversity can be mitigated by searching over candidate plans for solving a problem in natural language. Based on this insight, we propose PlanSearch, a novel search algorithm which shows strong results across HumanEval+, MBPP+, and LiveCodeBench (a contamination-free benchmark for competitive coding). PlanSearch generates a diverse set of observations about the problem and then uses these observations to construct plans for solving the problem. By searching over plans in natural language rather than directly over code solutions, PlanSearch explores a significantly more diverse range of potential solutions compared to baseline search methods. Using PlanSearch on top of Claude 3.5 Sonnet achieves a state-of-the-art pass@200 of 77.0% on LiveCodeBench, outperforming both the best score achieved without search (pass@1 = 41.4%) and using standard repeated sampling (pass@200 = 60.6%). Finally, we show that, across all models, search algorithms, and benchmarks analyzed, we can accurately predict performance gains due to search as a direct function of the diversity over generated ideas. Code can be found at https://github.com/scaleapi/plansearch.

In this paper, we introduce the VerifAI project, a pioneering open-source scientific question-answering system, designed to provide answers that are not only referenced but also automatically vetted and verifiable. The components of the system are (1) an Information Retrieval system combining semantic and lexical search techniques over scientific papers (PubMed), (2) a Retrieval-Augmented Generation (RAG) module using fine-tuned generative model (Mistral 7B) and retrieved articles to generate claims with references to the articles from which it was derived, and (3) a Verification engine, based on a fine-tuned DeBERTa and XLM-RoBERTa models on Natural Language Inference task using SciFACT dataset. The verification engine cross-checks the generated claim and the article from which the claim was derived, verifying whether there may have been any hallucinations in generating the claim. By leveraging the Information Retrieval and RAG modules, Verif.ai excels in generating factual information from a vast array of scientific sources. At the same time, the Verification engine rigorously double-checks this output, ensuring its accuracy and reliability. This dual-stage process plays a crucial role in acquiring and confirming factual information, significantly enhancing the information landscape. Our methodology could significantly enhance scientists' productivity, concurrently fostering trust in applying generative language models within scientific domains, where hallucinations and misinformation are unacceptable., Comment: Accepted at the 6th International Open Search Symposium 2024. arXiv admin note: substantial text overlap with arXiv:2402.18589

In this paper, we present a study of a Federated Learning (FL) system, based on the use of decentralized architectures to ensure trust and increase reliability. The system is based on the idea that the FL collaborators upload the (ciphered) model parameters on the Inter-Planetary File System (IPFS) and interact with a dedicated smart contract to track their behavior. Thank to this smart contract, the phases of parameter updates are managed efficiently, thereby strengthening data security. We have carried out an experimental study that exploits two different methods of weight aggregation, i.e., a classic averaging scheme and a federated proximal aggregation. The results confirm the feasibility of the proposal., Comment: TRUSTCHAIN workshop

Large language models (LLMs) have recently become the leading source of answers for users' questions online. Despite their ability to offer eloquent answers, their accuracy and reliability can pose a significant challenge. This is especially true for sensitive domains such as biomedicine, where there is a higher need for factually correct answers. This paper introduces a biomedical retrieval-augmented generation (RAG) system designed to enhance the reliability of generated responses. The system is based on a fine-tuned LLM for the referenced question-answering, where retrieved relevant abstracts from PubMed are passed to LLM's context as input through a prompt. Its output is an answer based on PubMed abstracts, where each statement is referenced accordingly, allowing the users to verify the answer. Our retrieval system achieves an absolute improvement of 23% compared to the PubMed search engine. Based on the manual evaluation on a small sample, our fine-tuned LLM component achieves comparable results to GPT-4 Turbo in referencing relevant abstracts. We make the dataset used to fine-tune the models and the fine-tuned models based on Mistral-7B-instruct-v0.1 and v0.2 publicly available., Comment: Accepted at BioNLP Workshop 2024, colocated with ACL 2024

Labelled tableaux have been a traditional approach to define satisfiability checking procedures for Modal Logics. In many cases, they can also be used to obtained tight complexity bounds and lead to efficient implementations of reasoning tools. More recently, it has been shown that the expressive power provided by the operators characterizing Hybrid Logics (nominals and satisfiability modalities) can be used to internalize labels, leading to well-behaved inference procedures for fairly expressive logics. The resulting procedures are attractive because they do not use external mechanisms outside the language of the logic at hand, and have good logical and computational properties. Lately, many proof systems based on Hybrid Logic have been investigated, in particular related to Modal Logics featuring some form of data comparison. In this paper, we introduce an internalized tableaux calculus for XPath, arguably one of the most prominent approaches for querying semistructured data. More precisely, we define data-aware tableaux for XPath featuring data comparison operators and enriched with nominals and the satisfiability modalities from Hybrid Logic. We prove that the calculus is sound, complete and terminating. Moreover, we show that tableaux can be constructed in polynomial space, without compromising completeness, establishing in this way that the satisfiability problem for the considered logic is PSPACE-complete. Finally, we explore different extensions of the calculus, in particular to handle data trees and other frame classes., Comment: 29 pages

A fraction of galaxy clusters harbor diffuse radio sources known as radio halos. The currently adopted scenario for their formation is based on second-order Fermi re-acceleration of seed electrons that is driven by merger-driven turbulence in the intra-cluster medium. This mechanism is expected to be inefficient, which implies that a significant fraction of halos should have very steep ($\alpha < -1.5$) energy spectra. We start investigating the potential and current limitations of the combination of the two surveys conducted by LOFAR, LoTSS (144 MHz) and LoLSS (54 MHz), to probe the origin of radio halos. We follow up the 20 radio halos detected in the DR1 of LoTSS, which covers the HETDEX field, with the LoLSS survey, and we study their spectral properties between 54 and 144 MHz. After the removal of compact sources, 9 halos were excluded due to unreliable halo flux density measurements at 54 MHz. Our main finding is that 7 out of 11 ($\sim$ 64%) exhibit an ultra-steep spectrum ($\alpha < -1.5$), which is a key prediction of turbulent re-acceleration models. We also note a tentative trend for more massive systems to host flatter halos, although the currently poor statistics does not allow for a deeper analysis. Our sample suffers from low angular resolution at 54 MHz, which limits the accuracy of the compact-sources subtraction. Nevertheless, this study is the first step towards providing compelling evidence for the existence of a large fraction of radio halos with very steep spectrum, which is a fundamental prediction of turbulent re-acceleration models. In this regard, the forthcoming second data release of LoLSS, along with the integration of LOFAR international stations and the instrumental upgrade to LOFAR2.0, will improve both the statistics and the low-frequency angular resolution, allowing to conclusively determine the origin of radio halos in galaxy clusters., Comment: A&A Letter; 11 pages, 12 figures

We introduce DafnyBench, the largest benchmark of its kind for training and evaluating machine learning systems for formal software verification. We test the ability of LLMs such as GPT-4 and Claude 3 to auto-generate enough hints for the Dafny formal verification engine to successfully verify over 750 programs with about 53,000 lines of code. The best model and prompting scheme achieved 68% success rate, and we quantify how this rate improves when retrying with error message feedback and how it deteriorates with the amount of required code and hints. We hope that DafnyBench will enable rapid improvements from this baseline as LLMs and verification techniques grow in quality., Comment: Code & dataset available at: https://github.com/sun-wendy/DafnyBench

We discuss Q-absolutely continuous functions between Carnot groups, following Maly's definition for maps of several variables. Such maps enjoy nice regularity properties, like continuity, Pansu differentiability a.e., weak differentiability and an Area formula. Furthermore, we extend Stein's result concerning the sharp condition for continuity and differentiability a.e. of a Sobolev map in terms of the integrability of the weak gradient: more precisely, we prove that a Sobolev map between Carnot groups with horizontal gradient of its sections uniformly bounded in L(Q,1) admits a representative which is Q-absolutely continuous.

Recent low-frequency radio observations at 140 MHz discovered a 3 Mpc-long bridge of diffuse emission connecting the galaxy clusters Abell 0399 and Abell 0401. We present follow-up observations at 60 MHz to constrain the spectral index of the bridge, which so far has only been detected at 140 and 144 MHz. We analysed deep (~18 hours) LOw Frequency ARray (LOFAR) Low Band Antenna (LBA) data at 60 MHz to detect the bridge at very low frequencies. We then conducted a multi-frequency study with LOFAR HBA data at 144 MHz and uGMRT data at 400 MHz. Assuming second-order Fermi mechanisms for the re-acceleration of relativistic electrons driven by turbulence in the radio bridge regions, we compare the observed radio spectrum with theoretical synchrotron models. The bridge is detected in the 75'' resolution LOFAR image at 60 MHz and its emission fully connects the region between the two galaxy clusters. Between 60 MHz and 144 MHz we found an integrated spectral index value of -1.44 +\- 0.16 for the bridge emission. For the first time, we produced spectral index and related uncertainties maps for a radio bridge. We produce a radio spectrum, which show significant steepening between 144 and 400 MHz. This detection at low frequencies provides important information on the models of particle acceleration and magnetic field structure on very extended scales. The spectral index gives important clues to the origin of inter-cluster diffuse emission. The steepening of the spectrum above 144 MHz can be explained in a turbulent re-acceleration framework, assuming that the acceleration timescales are longer than ~200 Myr., Comment: Letter to the Editor: 7 pages, 4 figures, 2 tables; accepted for publication in Astronomy & Astrophysics (A&A)

Classical schedulers for Battery Energy Storage Systems (BESSs) use static power constraints, assuming that the BESS can provide the rated power at any State-Of-Charge (SOC) level and that these are representative of the underlying physical constraints of the system (BESS voltage and current). Static power constraints, however, can generate unfeasible schedules, especially in power-intensive applications, as demonstrated in this paper. This paper derives a set of alternative constraints for the BESS power that are cognizant of the physical limits of the BESS. It is shown that these constraints, developed by leveraging an equivalent circuit model of the BESS, can be formulated as linear inequalities in scheduling problems, thus leaving the properties of the original problem (i.e., linearity and convexity) unaltered. A comparative analysis against traditional schedulers from the literature shows significant reductions in current violations and the generation of feasible schedules. These findings underscore the crucial role of implementing more advanced power constraints of BESSs in power-intensive applications, thereby enhancing the reliability of BESS scheduling strategies.

Radio observations of a few cool-core galaxy clusters have revealed the presence of diffuse emission on cluster scales, similar to what was found in merging clusters in the form of radio halos. These sources might suggest that a minor merger, while not sufficiently energetic to disrupt the cool core, could still trigger particle acceleration in the intracluster medium on scales of hundreds of kpc. We observed with LOFAR at 144 MHz a sample of twelve cool-core galaxy clusters presenting some level of dynamical disturbances, according to X-ray data. We also performed a systematic search of cold fronts in these clusters, re-analysing archival Chandra data. The clusters PSZ1G139.61+24, A1068 (new detection), MS 1455.0+2232, and RX J1720.1+2638 present diffuse radio emission on a cluster scale. This emission is characterised by a double component: a central mini-halo confined by cold fronts and diffuse emission on larger scales, whose radio power at 144 MHz is comparable to that of radio halos detected in merging systems. The cold fronts in A1068 are a new detection. We also found a candidate plasma depletion layer in this cluster. No sloshing features are found in the other eight clusters. Two of them present a mini-halo, with diffuse radio emission confined to the cluster core. We also found a new candidate mini-halo. Whereas, for the remaining five clusters, we did not detect halo-like emission. For clusters without cluster-scale halos, we derived upper limits to the radio halo power. We found that cluster-scale diffuse radio emission is not present in all cool-core clusters when observed at a low frequency, but it is correlated to the presence of cold fronts. This morphology requires a specific configuration of the merger and so it puts some constraints on the turbulence, which deserves to be investigated in the future with theoretical works., Comment: 34 pages, 30 figures. Accepted for publication in A&A

The BigCode project, an open-scientific collaboration focused on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder2. In partnership with Software Heritage (SWH), we build The Stack v2 on top of the digital commons of their source code archive. Alongside the SWH repositories spanning 619 programming languages, we carefully select other high-quality data sources, such as GitHub pull requests, Kaggle notebooks, and code documentation. This results in a training set that is 4x larger than the first StarCoder dataset. We train StarCoder2 models with 3B, 7B, and 15B parameters on 3.3 to 4.3 trillion tokens and thoroughly evaluate them on a comprehensive set of Code LLM benchmarks. We find that our small model, StarCoder2-3B, outperforms other Code LLMs of similar size on most benchmarks, and also outperforms StarCoderBase-15B. Our large model, StarCoder2- 15B, significantly outperforms other models of comparable size. In addition, it matches or outperforms CodeLlama-34B, a model more than twice its size. Although DeepSeekCoder- 33B is the best-performing model at code completion for high-resource languages, we find that StarCoder2-15B outperforms it on math and code reasoning benchmarks, as well as several low-resource languages. We make the model weights available under an OpenRAIL license and ensure full transparency regarding the training data by releasing the SoftWare Heritage persistent IDentifiers (SWHIDs) of the source code data.

The connection between the thermal and non-thermal properties in galaxy clusters hosting radio halos seems fairly well established. However, a comprehensive analysis of such a connection has been made only for integrated quantities (e.g. $L_X - P_{radio}$ relation). In recent years new-generation radio telescopes have enabled the unprecedented possibility to study the non-thermal properties of galaxy clusters on a spatially resolved basis. Here, we perform a pilot study to investigate the mentioned properties on five targets, by combining X-ray data from the CHEX-MATE project with the second data release from the LOFAR Two meter Sky survey. We find a strong correlation ($r_s \sim 0.7$) with a slope less than unity between the radio and X-ray surface brightness. We also report differences in the spatially resolved properties of the radio emission in clusters which show different levels of dynamical disturbance. In particular, less perturbed clusters (according to X-ray parameters) show peaked radio profiles in the centre, with a flattening in the outer regions, while the three dynamically disturbed clusters have steeper profiles in the outer regions. We fit a model to the radio emission in the context of turbulent re-acceleration with a constant ratio between thermal and non-thermal particles energy density and a magnetic field profile linked to the thermal gas density as $B(r) \propto n_{th}^{0.5}$. We found that this simple model cannot reproduce the behaviour of the observed radio emission., Comment: 23 pages, 10 figures, accepted for publication in A&A

Large Language Models (LLMs) can generate useful code, but often the code they generate cannot be trusted to be sound. In this paper, we present VerMCTS, an approach to begin to resolve this issue by generating verified programs in Dafny and Coq. VerMCTS uses a logical verifier in concert with an LLM to guide a modified Monte Carlo Tree Search (MCTS). This approach leverages the verifier to gain intermediate feedback inside the search algorithm by checking partial programs at each step to estimate an upper bound on the value function. To measure the performance of VerMCTS, we develop a new suite of multi-step verified programming problems in Dafny and Coq. In terms of pass@T, a new metric which computes the pass rate given a budget of T tokens sampled from the LLM, VerMCTS leads to more than a 30% absolute increase in average pass@5000 across the suite over repeated sampling from the base language model. Our code and benchmarks are available at https://github.com/namin/llm-verified-with-monte-carlo-tree-search .

Low-frequency radio observations show an increasing number of radio galaxies located in galaxy clusters that display peculiar morphologies and spectral profiles. This is the result of the dynamical interaction of the galaxy with the surrounding medium. Studying this phenomenon is key to understanding the evolution of low-energy relativistic particles in the intracluster medium. We present a multi-frequency study of the three head-tail (HT) radio galaxies and the radio halo in the galaxy cluster ZwCl0634.1+4747. We make use of observations at four frequencies performed with LOFAR LBA (53 MHz), HBA (144 MHz), GMRT (323 MHz) and VLA (1518 MHz) data. The use of extremely low radio frequency observations, such as LOFAR at 53 and 144 MHz, allowed us to detect the extension of the tails up to a distance of ~ 1 Mpc. We extracted spectral profiles along the tails in order to identify possible departures from a pure ageing model, such as the Jaffe-Perola (JP) model, which only involves synchrotron and inverse-Compton losses. We found clear evidence of departures from this simple ageing model, such as surface brightness enhancement and spectral flattening along all of the tails. This can be interpreted as the consequence of particle re-acceleration along the tails. Possible explanations for this behaviour include the interaction between a shock and the radio tails or a turbulence-driven re-acceleration mechanism. We show that the latter scenario is able to reproduce the characteristic features that we observed in our profiles.

A significant amount of research is focused on developing and evaluating large language models for a variety of code synthesis tasks. These include synthesizing code from natural language, synthesizing tests from code, and synthesizing explanations of code. In contrast, the behavior of instructional code editing with LLMs is understudied. These are tasks in which the model is provided a block of code and an instruction to modify the code. The editing instruction may ask for a feature to be added or removed, describe a bug and ask for a fix, or ask for a different kind of solution. We introduce a carefully crafted benchmark of code editing tasks and use it to evaluate several cutting edge LLMs. Our evaluation exposes a significant gap between the capabilities of state-of-the-art open and closed models. For example, even GPT-3.5-Turbo is better than the best open model at code editing tasks. We also introduce a new, carefully curated, permissively licensed training dataset of code editing tasks coupled with natural language instructions. Using this training dataset, we show that we can fine-tune open Code LLMs to significantly improve their code editing capabilities, closing the gap between open and closed models. All code, data, and models are available at https://github.com/nuprl/CanItEdit.

We present the first detailed analysis of the ultra-steep spectrum radio halo in the merging galaxy cluster Abell 521, based on upgraded Giant Metrewave Radio telescope (uGMRT) observations. The combination of radio observations (300-850 MHz) and archival X-ray data provide a new window into the complex physics occurring in this system. When compared to all previous analyses, our sensitive radio images detected the centrally located radio halo emission to a greater extent of $\sim$ 1.3 Mpc. A faint extension of the southeastern radio relic has been discovered. We detected another relic, recently discovered by MeerKAT, and coincident with a possible shock front in the X-rays, at the northwest position of the center. We find that the integrated spectrum of the radio halo is well-fitted with a spectral index of $-1.86 \pm 0.12$. A spatially resolved spectral index map revealed the spectral index fluctuations, as well as an outward radial steepening of the average spectral index. The radio and X-ray surface brightness are well correlated for the entire and different sub-parts of the halo, with sub-linear correlation slopes (0.50$-$0.65). We also found a mild anti-correlation between the spectral index and X-ray surface brightness. Newly detected extensions of the SE relic and the counter relic are consistent with the merger in the plane of the sky., Comment: 31 pages (total), 35 figures, 7 tables, 2 appendix (12 figures, 2 tables). Accepted for publication in Astrophysical Journal (APJ)

In this work, we investigated the interplay between the X-ray and radio emission of the cluster PSZ2G113.91-37.01 (z = 0.371) using the high-quality XMM-Newton observations of the CHEX-MATE project, and the images of the LoTSS-DR2. The cluster is undergoing a merger along the north-south axis, and shows a central radio halo and two radio relics, one in the southern and one in the northern regions. The analysis of the intracluster medium distribution revealed the presence of a northern surface brightness jump associated to the merger event. By extracting spectra across this discontinuity, we classified the edge as a cold front. Furthermore, we made use of upgraded Giant Metrewave Radio Telescope observations that allowed us to perform a spectral analysis of the G113 radio emission. We found evidence of re-acceleration of particles in the northern relic, and we measured an associated Mach number of M = 1.95 $\pm$ 0.01, as inferred from radio observations. We then performed a point-to-point analysis of the X-ray and radio emission both in the halo and in the northern relic regions. We found a strong correlation for the halo and an anti-correlation for the relic. The former behaviour is in agreement with previous studies. The relic anti-correlation is likely related to the reverse radial distribution of the X-ray (increasing towards the cluster centre) and radio (decreasing towards the cluster centre) emissions. Finally, we performed a point-to-point analysis of the radio emission and the residuals obtained by subtracting a double beta model to the X-ray emission. We found a strong correlation between the two quantities. This behaviour suggests the presence of a connection between the process responsible for the radio emission and the one that leaves fluctuations in the X-ray observations., Comment: 13 pages, 8 figures, 1 table. Accepted for publication in A&A

We investigate the complexity of the satisfiability problem for a modal logic expressing `knowing how' assertions, related to an agent's abilities to achieve a certain goal. We take one of the most standard semantics for this kind of logics based on linear plans. Our main result is a proof that checking satisfiability of a `knowing how' formula can be done in $\Sigma_2^P$. The algorithm we present relies on eliminating nested modalities in a formula, and then performing multiple calls to a satisfiability checking oracle for propositional logic.

Machine Learning (ML) is currently being exploited in numerous applications being one of the most effective Artificial Intelligence (AI) technologies, used in diverse fields, such as vision, autonomous systems, and alike. The trend motivated a significant amount of contributions to the analysis and design of ML applications against faults affecting the underlying hardware. The authors investigate the existing body of knowledge on Deep Learning (among ML techniques) resilience against hardware faults systematically through a thoughtful review in which the strengths and weaknesses of this literature stream are presented clearly and then future avenues of research are set out. The review is based on 220 scientific articles published between January 2019 and March 2024. The authors adopt a classifying framework to interpret and highlight research similarities and peculiarities, based on several parameters, starting from the main scope of the work, the adopted fault and error models, to their reproducibility. This framework allows for a comparison of the different solutions and the identification of possible synergies. Furthermore, suggestions concerning the future direction of research are proposed in the form of open challenges to be addressed., Comment: Submitted to Elsevier Computer Science Review on May 9, 2024

Software developers typically rely upon a large network of dependencies to build their applications. For instance, the NPM package repository contains over 3 million packages and serves tens of billions of downloads weekly. Understanding the structure and nature of packages, dependencies, and published code requires datasets that provide researchers with easy access to metadata and code of packages. However, prior work on NPM dataset construction typically has two limitations: 1) only metadata is scraped, and 2) packages or versions that are deleted from NPM can not be scraped. Over 330,000 versions of packages were deleted from NPM between July 2022 and May 2023. This data is critical for researchers as it often pertains to important questions of security and malware. We present npm-follower, a dataset and crawling architecture which archives metadata and code of all packages and versions as they are published, and is thus able to retain data which is later deleted. The dataset currently includes over 35 million versions of packages, and grows at a rate of about 1 million versions per month. The dataset is designed to be easily used by researchers answering questions involving either metadata or program analysis. Both the code and dataset are available at https://dependencies.science.

Over the past few years, Large Language Models of Code (Code LLMs) have started to have a significant impact on programming practice. Code LLMs are also emerging as building blocks for research in programming languages and software engineering. However, Code LLMs produce impressive results on programming languages that are well represented in their training data (e.g., Java, Python, or JavaScript), but struggle with low-resource languages that have limited training data available. Low resource languages include OCaml, Racket, and several others. This paper presents an effective approach for boosting the performance of Code LLMs on low-resource languages using semi-synthetic data. Our approach, MultiPL-T, translates training data from high-resource languages into training data for low-resource languages in the following way. 1) We use a Code LLM to synthesize tests for commented code from a high-resource language, filtering out faulty tests and code with low test coverage. 2) We use a Code LLM to translate Python code to a target low-resource language, and use tests to validate the translation. We apply this approach to generate tens of thousands of validated training items for Julia, Lua, OCaml, R, and Racket. Furthermore, we use an open model (StarCoderBase) with open training data (The Stack), which allows us to decontaminate benchmarks, train models without violating licenses, and run experiments that could not otherwise be done. With MultiPL-T generated data, we present fine-tuned versions of StarCoderBase and Code Llama for Julia, Lua, OCaml, R, and Racket. On established benchmarks (MultiPL-E), these models outperform other open Code LLMs. The MultiPL-T approach is easy to apply to new languages, and is significantly more efficient and effective than alternatives such as training longer.

Turbulence introduced into the intra-cluster medium (ICM) through cluster merger events transfers energy to non-thermal components, and can trigger the formation of diffuse synchrotron radio sources. Typical diffuse sources in the forms of giant radio halos and mini-halos are found in merging and relaxed cool core galaxy clusters, respectively. On the other hand, recent observations have revealed an increasing complexity of the non-thermal phenomenology. Abell 2142 (A2142) is a mildly disturbed cluster that exhibits uncommon thermal and non-thermal properties. It is known to host a hybrid halo consisting of two components (H1 and H2), namely a mini-halo-like and an enigmatic elongated radio halo-like structure. We aim to investigate the properties, origin, and connections of each component. We present deep LOFAR observations of A2142 in the frequency ranges $30-78$ MHz and $120-168$ MHz. With complementary multi-frequency radio and X-ray data, we analyse the radio spectral properties of the halo and assess the connection between the non-thermal and thermal components of the ICM. We detected a third radio component (H3), which extends over the cluster volume on scales $\sim 2$ Mpc, embeds H1 and H2, and has a morphology that roughly follows the thermal ICM distribution. The radio spectral index is moderately steep in H1 ($\alpha=1.09\pm 0.02$) and H2 ($\alpha=1.15\pm 0.02$), but is steeper ($\alpha=1.57\pm 0.20$) in H3. The analysis of the thermal and non-thermal properties allowed us to discuss possible formation scenarios for each radio component. Turbulence from sloshing motions of low-entropy gas on different scales may be responsible for the origin of H1 and H2. We classified H3 as a giant ultra-steep spectrum radio halo, which could trace the residual activity from an old energetic merger and/or inefficient turbulent re-acceleration induced by ongoing minor mergers., Comment: 24 pages (including Appendix), 15 Figures (plus 4 Figures in Appendix), Accepted for publication by A&A

The relation between giant radio halos and mini-halos in galaxy clusters is not understood. The former are usually associated with merging clusters, the latter are found in relaxed systems. In the last years, the advent of low-frequency radio observations has challenged this dichotomy, finding intermediate objects with a hybrid radio morphology. We aim to investigate the presence of diffuse radio emission in the cluster Abell 1413 and determine its dynamical status. We used LOFAR HBA observations centred at 144 MHz to study the diffuse emission hosted by this cluster.To investigate the dynamical state of the system, we complete our study with newly analysed XMM-Newton archival data. A1413 shows features that are typically present in both relaxed (e.g., peaked x-ray surface brightness distribution and little large-scale inhomogeneities) and disturbed (e.g., flatter temperature and metallicity profiles) clusters.This evidence supports the scenario that A1413 is neither a disturbed nor fully relaxed object. We argue that it is an intermediate-phase cluster.Using radio observations at 144 MHz, we discover the presence of a wider diffuse component surrounding the previously reported mini-halo at the cluster centre. By fitting the radio surface brightness profile with a double-exponential model, we can disentangle the two components. We find an inner mini-halo with an e-folding radius r_e1=28 kpc and the extended component with r_e2 = 290 kpc. We also performed point-to-point correlations between radio and X-ray surface brightness, finding a sub-linear relation for the outer emission and a super-linear relation for the mini-halo.The mini-halo and the diffuse emission extend over different scales and show different features, confirming the double nature of the radio emission and suggesting that the mechanisms responsible for the re-acceleration of the radio-emitting particle might be different., Comment: 14 pages, 8 figures, accepted A&A

Many relaxed cool-core clusters host diffuse radio emission on scales of hundreds of kiloparsecs: mini-haloes. However, the mechanism responsible for generating them, as well as their connection with central active galactic nuclei, is elusive and many questions related to their physical properties and origins remain unanswered. This paper presents new radio observations of the galaxy cluster Abell 1413 performed with MeerKAT (L-band; 872 to 1712 MHz) and LOFAR HBA (120 to 168 MHz) as part of a statistical and homogeneous census of mini-haloes. Abell 1413 is unique among mini-halo clusters as it is a moderately-disturbed non-cool-core cluster. Our study reveals an asymmetric mini-halo up to 584 kpc in size at 1283 MHz, twice as large as first reported at similar frequencies. The spectral index is flatter than previously reported, with an integrated value of $\alpha = -1.01 \pm 0.06$, shows significant spatial variation, and a tentative radial steepening. We studied the point-to-point X-ray/radio surface brightness correlation to investigate the thermal/non-thermal connection: our results show a strong connection between these components, with a super-linear slope of $b = 1.63 \pm 0.10$ at 1283 MHz and $b = 1.20 \pm 0.12$ at 145 MHz. We also explore the X-ray surface brightness/radio spectral index correlation, finding a slope of $b = 0.59 \pm 0.11$. Both investigations support the evidence of spectral steepening. Finally, in the context of understanding the particle acceleration mechanism, we present a simple theoretical model which demonstrates that hybrid scenarios - secondary electrons (re-)accelerated by turbulence - reproduce a super-linear correlation slope., Comment: 19 pages, 11 figures, 4 tables. Accepted for publication in MNRAS

In this paper, we investigate properties of unique continuation for hyperbolic Schr\"odinger equations with time-dependent complex-valued electric fields and time-independent real magnetic fields. We show that positive masses inside of a bounded region at a single time propagate outside the region and prove gaussian lower bounds for the solutions, provided a suitable average in space-time cylinders is taken., Comment: arXiv admin note: substantial text overlap with arXiv:2107.10782