Search

Question answering is a natural language understanding task that involves reasoning over both explicit context and unstated, relevant domain knowledge. Large language models (LLMs), which underpin most contemporary question answering systems, struggle to induce how concepts relate in specialized domains such as medicine. Existing medical LLMs are also costly to train. In this work, we present MEG, a parameter-efficient approach for medical knowledge-augmented LLMs. MEG uses a lightweight mapping network to integrate graph embeddings into the LLM, enabling it to leverage external knowledge in a cost-effective way. We evaluate our method on four popular medical multiple-choice datasets and show that LLMs greatly benefit from the factual grounding provided by knowledge graph embeddings. MEG attains an average of +10.2% accuracy over the Mistral-Instruct baseline, and +6.7% over specialized models like BioMistral. We also show results based on Llama-3. Finally, we show that MEG's performance remains robust to the choice of graph encoder.

Bipartite perfect quantum strategies (BPQSs) allow two players isolated from each other to win every trial of a nonlocal game. BPQSs have crucial roles in recent developments in quantum information and quantum computation. However, only few BPQSs with a small number of inputs are known and only one of them has been experimentally tested. It has recently been shown that every BPQS has an associated Kochen-Specker (KS) set. Here, we first prove that any BPQS of minimum input cardinality that can be obtained from a generalized KS set can also be obtained from a KS set of pure states. Then, we address the problem of finding BPQSs of small input cardinality starting from KS sets. We introduce an algorithm that identifies the BPQS with the minimum number of settings for any given KS set. We apply it to many well-known KS sets of small input cardinality in dimensions $3$, $4$, $5$, $6$, $7$, and $8$. In each dimension, the algorithm either obtains the best BPQS known or find one with fewer inputs., Comment: 18 pages, 1 figure, 10 tables

Robots are frequently tasked to gather relevant sensor data in unknown terrains. A key challenge for classical path planning algorithms used for autonomous information gathering is adaptively replanning paths online as the terrain is explored given limited onboard compute resources. Recently, learning-based approaches emerged that train planning policies offline and enable computationally efficient online replanning performing policy inference. These approaches are designed and trained for terrain monitoring missions assuming a single specific map representation, which limits their applicability to different terrains. To address these issues, we propose a novel formulation of the adaptive informative path planning problem unified across different map representations, enabling training and deploying planning policies in a larger variety of monitoring missions. Experimental results validate that our novel formulation easily integrates with classical non-learning-based planning approaches while maintaining their performance. Our trained planning policy performs similarly to state-of-the-art map-specifically trained policies. We validate our learned policy on unseen real-world terrain datasets., Comment: 8 pages, 4 figures

We show that if $p>1$ every subspace of $\ell_p(\Gamma)$ is an $\ell_p$-sum of separable subspaces of $\ell_p$, and we provide examples of subspaces of $\ell_p(\Gamma)$ for $0

BL Lac objects are a class of jetted active galactic nuclei that do not exhibit or have weak emission lines in their optical spectra. Recently, the first $\gamma$-ray emitting BL Lac beyond $z=3$, 4FGL J1219.0 +3653 (hereafter J1219), was identified, i.e., within the first two billion years of the age of the universe. Here we report the results obtained from a detailed broadband study of this peculiar source by analyzing the new $\sim$58 ksec XMM-Newton and archival observations and reproducing the multiwavelength spectral energy distribution with the conventional one-zone leptonic radiative model. The XMM-Newton~data revealed that J1219 is a faint X-ray emitter ($F_{\rm 0.3-10~keV}=8.39^{+4.11}_{-2.40}\times10^{-15}$ erg/cm2/s) and exhibits a soft spectrum (0.3$-$10 keV photon index$=2.28^{+0.58}_{-0.48}$). By comparing the broadband physical properties of J1219 with $z>3$ $\gamma$-ray detected flat spectrum radio quasars (FSRQs), we have found that it has a relatively low jet power and, similar to FSRQs, the jet power is larger than the accretion disk luminosity. We conclude that deeper multiwavelength observations will be needed to fully explore the physical properties of this unique high-redshift BL Lac object., Comment: JHEAP, in press

Photonic memory is an important building block to delay, route and buffer optical information, for instance in optical interconnects or for recurrent optical signal processing. Photonic-phononic memory based on stimulated Brillouin-Mandelstam scattering (SBS) has been demonstrated as a coherent optical storage approach with broad bandwidth, frequency selectivity and intrinsic nonreciprocity. Here, we experimentally demonstrated the storage of quadrature-phase encoded data at room temperature and at cryogenic temperatures. We store and retrieve the 2-bit states $\{00, 01, 10, 11\}$ encoded as optical pulses with the phases $\{0, {\pi}/2 , {\pi}, 3{\pi}/2\}$ - a quadrature phase shift keying (QPSK) signal. The 2-bit signals are retrieved from the acoustic domain with a global phase rotation of ${\pi}$, which is inherent in the process due to SBS. We also demonstrate full phase control over the retrieved data based on two different handles: by detuning slightly from the SBS resonance, or by changing the storage time in the memory scheme we can cover the full range $[0, 2{\pi})$. At a cryogenic temperature of 3.9 K, we have increased readout efficiency as well as gained access to longer storage times, which results in a detectable signal at 140 ns. All in all, the work sets the cornerstone for optoacoustic memory schemes with phase-encoded data, Comment: O.S. and J.H.M.C. contributed equally to this work

In quantum cryptography, secure randomness expansion involves using a short private string of random bits to generate a longer one, even in the presence of an adversary who may have access to quantum resources. In this work, we demonstrate that local contextuality-based self-tests are sufficient to construct a randomness expansion protocol that is secure against computationally unbounded quantum adversaries. Our protocol is based on self-testing from non-contextuality inequalities and we prove that our scheme asymptotically produces secure random numbers which are $\mathcal{O}(m\sqrt{\epsilon})$-close to uniformly distributed and private, where $\epsilon$ is the robustness parameter of the self-test and $m$ is the length of the generated random bit string. Our protocol is semi-device-independent in the sense that it inherits any assumptions necessary for the underlying self-test., Comment: 5+12 pages. Comments are welcome!

Post-quantum cryptography has gained attention due to the need for secure cryptographic systems in the face of quantum computing. Code-based and lattice-based cryptography are two prominent approaches, both heavily studied within the NIST standardization project. Code-based cryptography -- most prominently exemplified by the McEliece cryptosystem -- is based on the hardness of decoding random linear error-correcting codes. Despite the McEliece cryptosystem having been unbroken for several decades, it suffers from large key sizes, which has led to exploring variants using metrics than the Hamming metric, such as the Lee metric. This alternative metric may allow for smaller key sizes, but requires further analysis for potential vulnerabilities to lattice-based attack techniques. In this paper, we consider a generic Lee metric based McEliece type cryptosystem and evaluate its security against lattice-based attacks.

Correctly assessing the malignancy of breast lesions identified during ultrasound examinations is crucial for effective clinical decision-making. However, the current "golden standard" relies on manual BI-RADS scoring by clinicians, often leading to unnecessary biopsies and a significant mental health burden on patients and their families. In this paper, we introduce PersonalizedUS, an interpretable machine learning system that leverages recent advances in conformal prediction to provide precise and personalized risk estimates with local coverage guarantees and sensitivity, specificity, and predictive values above 0.9 across various threshold levels. In particular, we identify meaningful lesion subgroups where distribution-free, model-agnostic conditional coverage holds, with approximately 90% of our prediction sets containing only the ground truth in most lesion subgroups, thus explicitly characterizing for which patients the model is most suitably applied. Moreover, we make available a curated tabular dataset of 1936 biopsied breast lesions from a recent observational multicenter study and benchmark the performance of several state-of-the-art learning algorithms. We also report a successful case study of the deployed system in the same multicenter context. Concrete clinical benefits include up to a 65% reduction in requested biopsies among BI-RADS 4a and 4b lesions, with minimal to no missed cancer cases., Comment: 9 pages, 5 figure, 2 tables

We study the problem of searching for a target at some unknown location in $\mathbb{R}^d$ when additional information regarding the position of the target is available in the form of predictions. In our setting, predictions come as approximate distances to the target: for each point $p\in \mathbb{R}^d$ that the searcher visits, we obtain a value $\lambda(p)$ such that $|p\mathbf{t}|\le \lambda(p) \le c\cdot |p\mathbf{t}|$, where $c\ge 1$ is a fixed constant, $\mathbf{t}$ is the position of the target, and $|p\mathbf{t}|$ is the Euclidean distance of $p$ to $\mathbf{t}$. The cost of the search is the length of the path followed by the searcher. Our main positive result is a strategy that achieves $(12c)^{d+1}$-competitive ratio, even when the constant $c$ is unknown. We also give a lower bound of roughly $(c/16)^{d-1}$ on the competitive ratio of any search strategy in $\mathbb{R}^d$., Comment: 19 pages, WAOA 24

A perfect strategy is one that allows the mutually in-communicated players of a nonlocal game to win every trial of the game. Perfect strategies are basic tools for some fundamental results in quantum computation and crucial resources for some applications in quantum information. Here, we address the problem of producing qudit-qudit perfect quantum strategies with a small number of settings. For that, we exploit a recent result showing that any perfect quantum strategy induces a Kochen-Specker set. We identify a family of KS sets in even dimension $d \ge 6$ that, for many dimensions, require the smallest number of orthogonal bases known: $d+1$. This family was only defined for some $d$. We first extend the family to infinitely many more dimensions. Then, we show the optimal way to use each of these sets to produce a bipartite perfect strategy with minimum input cardinality. As a result, we present a family of perfect quantum strategies in any $(2,d-1,d)$ Bell scenario, with $d = 2^kp^m$ for $p$ prime, $m \geq k \geq 0$ (excluding $m=k=0$), $d = 8p$ for $p \geq 19$, $d=kp$ for $p > ((k-2)2^{k-2})^2$ whenever there exists a Hadamard matrix of order $k$, other sporadic examples, as well as a recursive construction that produces perfect quantum strategies for infinitely many dimensions $d$ from any dimension $d'$ with a perfect quantum strategy. We identify their associated Bell inequalities and prove that they are not tight, which provides a second counterexample to a conjecture of 2007., Comment: 26 pages, 5 figures

We prove the Mond conjecture for wave fronts which states that the number of parameters of a frontal versal unfolding is less than or equal to the number of spheres in the image of a stable frontal deformation with equality if the wave front is weighted homogeneous. We give two different proofs. The first one depends on the fact that wave fronts are related to discriminants of map germs and we then use the analogous result proved by Damon and Mond in this context. The second one is based on ideas by Fern\'andez de Bobadilla, Nu\~no-Ballesteros and Pe\~nafort Sanchis and by Nu\~no-Ballesteros and Fern\'andez-Hern\'andez. The advantage of the second approach is that most results are valid for any frontal, not only wave fronts, and thus give important tools which may be useful to prove the conjecture for frontals in general., Comment: 21 pages, 4 figures

We show that each set of $n\ge 2$ points in the plane in general position has a straight-line matching with at least $(5n+1)/27$ edges whose segments form a connected set, and such a matching can be computed in $O(n \log n)$ time. As an upper bound, we show that for some planar point sets in general position the largest matching whose segments form a connected set has $\lceil \frac{n-1}{3}\rceil$ edges. We also consider a colored version, where each edge of the matching should connect points with different colors., Comment: 20 pages, 14 figures; preliminary version in EuroCG 2024

We report the first experimental demonstration of the odd-cycle game. We entangle two ions separated by ~2 m and the players use them to win the odd-cycle game with a probability ~26 sigma above that allowed by the best classical strategy. The experiment implements the optimal quantum strategy, is free of loopholes, and achieves 97.8(3) % of the theoretical limit to the quantum winning probability. We perform the associated Bell test and measure a nonlocal content of 0.54(2) -- the largest value for physically separate devices, free of the detection loophole, ever observed.

We show that quantum computers can be used for producing large $n$-partite nonlocality, thereby providing a method to benchmark them. The main challenges to overcome are: (i) The interaction topology might not allow arbitrary two-qubit gates. (ii) Noise limits the Bell violation. (iii) The number of combinations of local measurements grows exponentially with $n$. To overcome (i), we point out that graph states that are compatible with the two-qubit connectivity of the computer can be efficiently prepared. To mitigate (ii), we note that, for specific graph states, there are $n$-partite Bell inequalities whose resistance to white noise increases exponentially with $n$. To address (iii) for any $n$ and any connectivity, we introduce an estimator that relies on random sampling. As a result, we propose a method for producing $n$-partite Bell nonlocality with unprecedented large $n$. This allows in return to benchmark nonclassical correlations regardless of the number of qubits or the connectivity. We test our approach by using a simulation for a noisy IBM quantum computer, which predicts $n$-partite Bell nonlocality for at least $n=24$ qubits., Comment: 14 pages, 13 figures; Improvements in the presentation and the noise model

Aspect-Based Sentiment Analysis (ABSA) involves extracting opinions from textual data about specific entities and their corresponding aspects through various complementary subtasks. Several prior research has focused on developing ad hoc designs of varying complexities for these subtasks. In this paper, we present a generative framework extensible to any ABSA subtask. We build upon the instruction tuned model proposed by Scaria et al. (2023), who present an instruction-based model with task descriptions followed by in-context examples on ABSA subtasks. We propose PFInstruct, an extension to this instruction learning paradigm by appending an NLP-related task prefix to the task description. This simple approach leads to improved performance across all tested SemEval subtasks, surpassing previous state-of-the-art (SOTA) on the ATE subtask (Rest14) by +3.28 F1-score, and on the AOOE subtask by an average of +5.43 F1-score across SemEval datasets. Furthermore, we explore the impact of the prefix-enhanced prompt quality on the ABSA subtasks and find that even a noisy prefix enhances model performance compared to the baseline. Our method also achieves competitive results on a biomedical domain dataset (ERSA)., Comment: Accepted to ACL 2024 Findings

Drawing a graph in the plane with as few crossings as possible is one of the central problems in graph drawing and computational geometry. Another option is to remove the smallest number of vertices or edges such that the remaining graph can be drawn without crossings. We study both problems in a book-embedding setting for ordered graphs, that is, graphs with a fixed vertex order. In this setting, the vertices lie on a straight line, called the spine, in the given order, and each edge must be drawn on one of several pages of a book such that every edge has at most a fixed number of crossings. In book embeddings, there is another way to reduce or avoid crossings; namely by using more pages. The minimum number of pages needed to draw an ordered graph without any crossings is its (fixed-vertex-order) page number. We show that the page number of an ordered graph with $n$ vertices and $m$ edges can be computed in $2^m \cdot n^{O(1)}$ time. An $O(\log n)$-approximation of this number can be computed efficiently. We can decide in $2^{O(d \sqrt{k} \log (d+k))} \cdot n^{O(1)}$ time whether it suffices to delete $k$ edges of an ordered graph to obtain a $d$-planar layout (where every edge crosses at most $d$ other edges) on one page. As an additional parameter, we consider the size $h$ of a hitting set, that is, a set of points on the spine such that every edge, seen as an open interval, contains at least one of the points. For $h=1$, we can efficiently compute the minimum number of edges whose deletion yields fixed-vertex-order page number $p$. For $h>1$, we give an XP algorithm with respect to $h+p$. Finally, we consider spine+$t$-track drawings, where some but not all vertices lie on the spine. The vertex order on the spine is given; we must map every vertex that does not lie on the spine to one of $t$ tracks, each of which is a straight line on a separate page, parallel to the spine., Comment: Appears in Proc. 19th Scandinavian Symposium on Algorithm Theory (SWAT 2024)

Anatomically guided PET reconstruction using MRI information has been shown to have the potential to improve PET image quality. However, these improvements are limited to PET scans with paired MRI information. In this work we employed a diffusion probabilistic model (DPM) to infer T1-weighted-MRI (deep-MRI) images from FDG-PET brain images. We then use the DPM-generated T1w-MRI to guide the PET reconstruction. The model was trained with brain FDG scans, and tested in datasets containing multiple levels of counts. Deep-MRI images appeared somewhat degraded than the acquired MRI images. Regarding PET image quality, volume of interest analysis in different brain regions showed that both PET reconstructed images using the acquired and the deep-MRI images improved image quality compared to OSEM. Same conclusions were found analysing the decimated datasets. A subjective evaluation performed by two physicians confirmed that OSEM scored consistently worse than the MRI-guided PET images and no significant differences were observed between the MRI-guided PET images. This proof of concept shows that it is possible to infer DPM-based MRI imagery to guide the PET reconstruction, enabling the possibility of changing reconstruction parameters such as the strength of the prior on anatomically guided PET reconstruction in the absence of MRI.

Advancements in instrumentation have revealed a multitude of small-scale EUV events in the solar atmosphere. Our aim is to employ high-resolution magnetograms to gain a detailed understanding of the magnetic origin of such phenomena. We have used coordinated observations from SST, IRIS, and SDO to analyze an ephemeral magnetic flux emergence episode and the following chain of small-scale energetic events. These unique observations clearly link these phenomena together. The high-resolution (0."057/pixel) magnetograms obtained with SST/CRISP allows us to reliably measure the magnetic field at the photosphere and detect the emerging bipole that causes the subsequent eruptive atmospheric events. Notably, this small-scale emergence episode remains indiscernible in the lower resolution SDO/HMI magnetograms (0."5/pixel). We report the appearance of a dark bubble in Ca II K related to the emerging bipole, a sign of the canonical expanding magnetic dome predicted in flux emergence simulations. Evidences of reconnection are also found: first through an Ellerman bomb, and later by the launch of a surge next to a UV burst. The UV burst exhibits a weak EUV counterpart in the coronal SDO/AIA channels. By calculating DEM, its plasma is shown to reach a temperature beyond 1 MK and have densities between the upper chromosphere and transition region. Our study showcases the importance of high-resolution magnetograms to unveil the mechanisms triggering phenomena such as EBs, UV bursts, and surges. This could hold implications for small-scale events akin to those recently reported in EUV using Solar Orbiter. The finding of temperatures beyond 1 MK in the UV burst plasma strongly suggests that we are examining analogous features. Therefore, we signal caution regarding drawing conclusions from full-disk magnetograms that lack the necessary resolution to reveal their true magnetic origin., Comment: Accepted in A&A, 11 pages, 7 figures, 5 movies

Since their introduction in 1997, the Hardy spaces of Dirichlet series have been broadly and deeply studied. The increasing interest sparked by these Banach spaces of Dirichlet series motivated the introduction of new such spaces, as the Bergman spaces of Dirichlet series $\mathcal{A}^p_{\mu}$ here considered, where $\mu$ is a probability measure on $(0,\infty)$. Similarly, recent lines of research have focused their attention on the study of some classical operators acting on these spaces, as it is the case of the Volterra operator $T_g$. In this work, we introduce a new family of Bloch spaces of Dirichlet series, the $\text{Bloch}_{\mu}$-spaces, and study some of its most essential properties. Using these spaces we are able to provide a sufficient condition for the Volterra operator $T_g$ to act boundedly on the Bergman spaces $\mathcal{A}^p_{\mu}$. We also establish a necessary condition for a specific choice of the probability measures $\mu$. Sufficient and necessary conditions for compactness are also proven. The membership in Schatten classes is studied as well. Eventually, a radicality result is established for Bloch spaces of Dirichlet series., Comment: Main changes in Section 3 and Section 7. Some results have been improved and some proofs have been written more clearly

The theory of Banach spaces of Dirichlet series has drawn an increasing attention in the recent 25 years. One of the main interest of this new theory is that of defining analogues of the classical spaces of analytic functions on the unit disc. In this sense, Bergman spaces were introduced several years ago contributing to broaden the picture of this theory. In the paper presenting these new family of spaces, some of its most essential questions were considered. Among them, some partial estimates of the norm of the pointwise evaluation functional were given. In this work, we introduce a version of the Riemann-Liouville semigroup acting on these spaces, and, with this new tool, we are able to estimate the norm of this functional., Comment: Somo general corrections in order to ease the reading and the statements of the results

Our main result states that whenever we have a non-Euclidean norm $\|\cdot\|$ on a two-dimensional vector space $X$, there exists some $x\neq 0$ such that for every $\lambda\neq 1, \lambda>0$, there exist $y, z\in X$ verifying that $\|y\|=\lambda\|x\|$, $z\neq 0$, and $z$ belongs to the bisectors $B(-x,x)$ and $B(-y,y)$. Throughout this paper we also state and prove some other simple but maybe useful results about the geometry of the unit sphere of strictly convex planes.

There are a couple of purposes in this paper: to study a problem of approximation with exponential functions and to show its relevance for the economic science. We present results that completely solve the problem of the best approximation by means of exponential functions and we will be able to determine what kind of data is suitable to be fitted. Data will be approximated using TAC (implemented in the R-package nlstac), a numerical algorithm for fitting data by exponential patterns without initial guess designed by the authors. We check one more time the robustness of this algorithm by successfully applying it to two very distant areas of economy: demand curves and nonlinear time series. This shows TAC's utility and highlights how far this algorithm could be used.

A new package for nonlinear least squares fitting is introduced in this paper. This package implements a recently developed algorithm that, for certain types of nonlinear curve fitting, reduces the number of nonlinear parameters to be fitted. One notable feature of this method is the absence of initialization which is typically necessary for nonlinear fitting gradient-based algorithms. Instead, just some bounds for the nonlinear parameters are required. Even though convergence for this method is guaranteed for exponential decay using the max-norm, the algorithm exhibits remarkable robustness, and its use has been extended to a wide range of functions using the Euclidean norm. Furthermore, this data-fitting package can also serve as a valuable resource for providing accurate initial parameters to other algorithms that rely on them.

In this note we show that the usual notion of fuzzy norm defined on a linear space is equivalent to that of quasiconcave function, in the sense that every fuzzy norm $N:X\times\mathbb{R}[0,1]$ defined on a (real or complex) linear space X is uniquely determined by a quasiconcave function $f:X\to[0, 1]$. We explore the minimum requirements that we need to impose to some quasiconcave function $f:X\to[0, 1]$ in order to define a fuzzy norm $N:X\times\mathbb{R}[0,1]$. Later we use this equivalence to prove some properties of fuzzy norms, like a generalisation of the celebrated Decomposition Theorem.

The experimental violation of a Bell inequality implies that at least one of a set of assumptions fails in nature. However, existing tests are inconclusive about which of the assumptions is the one that fails. Here, we show that there are quantum correlations that cannot be simulated with hidden variables that allow the slightest free will (or, equivalently, that limit, even minimally, retrocausal influences) or restrict, even minimally, actions at a distance. This result goes beyond Bell's theorem and demolishes the arguably most attractive motivation for considering hidden-variable theories with measurement dependence or actions at distance, namely, that simulating quantum correlations typically requires a small amount of these resources. We show that there is a feasible experiment that can discard any hidden-variable theory allowing for arbitrarily small free will and having arbitrarily small limitations to actions at a distance. The experiment involves two observers, each of them choosing between two measurements with $2^N$ outcomes. The larger $N$ for which a specific Bell-like inequality is violated, the larger the set of excluded hidden-variable theories. In the limit of $N$ tending to infinity, the only alternatives to the absence of hidden variables are complete superdeterminism or complete parameter dependence. We also explore the implications of this result for quantum information., Comment: 6+18 pages, 1 figure, 1 table

Coronary artery disease (CAD) remains the leading cause of death globally and invasive coronary angiography (ICA) is considered the gold standard of anatomical imaging evaluation when CAD is suspected. However, risk evaluation based on ICA has several limitations, such as visual assessment of stenosis severity, which has significant interobserver variability. This motivates to development of a lesion classification system that can support specialists in their clinical procedures. Although deep learning classification methods are well-developed in other areas of medical imaging, ICA image classification is still at an early stage. One of the most important reasons is the lack of available and high-quality open-access datasets. In this paper, we reported a new annotated ICA images dataset, CADICA, to provide the research community with a comprehensive and rigorous dataset of coronary angiography consisting of a set of acquired patient videos and associated disease-related metadata. This dataset can be used by clinicians to train their skills in angiographic assessment of CAD severity and by computer scientists to create computer-aided diagnostic systems to help in such assessment. In addition, baseline classification methods are proposed and analyzed, validating the functionality of CADICA and giving the scientific community a starting point to improve CAD detection.

Kochen-Specker contextuality is a fundamental feature of quantum mechanics and a crucial resource for quantum computational advantage and reduction of communication complexity. Its presence is witnessed in empirical data by the violation of noncontextuality inequalities. However, all known noncontextuality inequalities corresponding to facets of noncontextual polytopes are either Bell inequalities or refer to cyclic or state-independent contextuality scenarios. We introduce a general method for lifting noncontextuality inequalities, deriving facets of noncontextual polytopes for more complex scenarios from known facets of simpler subscenarios. Concretely, starting from an arbitrary scenario, the addition of a new measurement or a new outcome preserves the facet-defining nature of any noncontextuality inequality. This extends the results of Pironio [J. Math. Phys. 46, 062112 (2005)] from Bell nonlocality scenarios to contextuality scenarios, unifying liftings of Bell and noncontextuality inequalities. Our method produces facet-defining noncontextuality inequalities in all scenarios with contextual correlations, and we present examples of facet-defining noncontextuality inequalities for scenarios where no examples were known. Our results shed light on the structure of noncontextuality polytopes and the relationship between such polytopes across different scenarios., Comment: 20 pages, 4 figures

Let $S$ be a set of four points chosen independently, uniformly at random from a square. Join every pair of points of $S$ with a straight line segment. Color these edges red if they have positive slope and blue, otherwise. We show that the probability that $S$ defines a pair of crossing edges of the same color is equal to $1/4$. This is connected to a recent result of Aichholzer et al. [GD 2019] who showed that by 2-colouring the edges of a geometric graph and counting monochromatic crossings instead of crossings, the number of crossings can be more than halfed. Our result shows that for the described random drawings, there is a coloring of the edges such that the number of monochromatic crossings is in expectation $\frac{1}{2}-\frac{7}{50}$ of the total number of crossings.

A bipartite perfect quantum strategy (BPQS) allows two players who cannot communicate with each other to always win a nonlocal game. BPQSs are rare but fundamental in light of some recent results in quantum information, computation, and foundations. A more than 40-year old open problem is how many inputs (measurement settings) a BPQS requires. A related problem is how many inputs are needed if, in addition, the quantum system has minimum dimension. A third, apparently unrelated, problem is what is the connection between BPQSs and state-independent contextuality, which inspired the first BPQSs. Here, we solve the third problem: We prove that {\em every} BPQS defines a Kochen-Specker (KS) set. We use this result to identify the BPQS with the smallest number of inputs, both in the general case and in the case of minimum dimension, and solve some related problems. We argue that either the BPQSs presented are the solutions to the first and second problems or, after more than 56 years of research, we have missed some important KS set., Comment: 6 pages, 2 figures

The majority of massive disk galaxies in the local Universe show a stellar barred structure in their central regions, including our Milky Way. Bars are supposed to develop in dynamically cold stellar disks at low redshift, as the strong gas turbulence typical of disk galaxies at high redshift suppresses or delays bar formation. Moreover, simulations predict bars to be almost absent beyond $z = 1.5$ in the progenitors of Milky Way-like galaxies. Here we report observations of ceers-2112, a barred spiral galaxy at redshift $z_{\rm phot} \sim 3$, which was already mature when the Universe was only 2 Gyr old. The stellar mass ($M_{\star} = 3.9 \times 10^9 M_{\odot}$) and barred morphology mean that ceers-2112 can be considered a progenitor of the Milky Way, in terms of both structure and mass-assembly history in the first 2 Gyr of the Universe, and was the closest in mass in the first 4 Gyr. We infer that baryons in galaxies could have already dominated over dark matter at $z \sim 3$, that high-redshift bars could form in approximately 400 Myr and that dynamically cold stellar disks could have been in place by redshift $z = 4-5$ (more than 12 Gyrs ago)., Comment: Published in Nature. Accepted on September 12

Pretrained machine learning models are known to perpetuate and even amplify existing biases in data, which can result in unfair outcomes that ultimately impact user experience. Therefore, it is crucial to understand the mechanisms behind those prejudicial biases to ensure that model performance does not result in discriminatory behaviour toward certain groups or populations. In this work, we define gender bias as our case study. We quantify bias amplification in pretraining and after fine-tuning on three families of vision-and-language models. We investigate the connection, if any, between the two learning stages, and evaluate how bias amplification reflects on model performance. Overall, we find that bias amplification in pretraining and after fine-tuning are independent. We then examine the effect of continued pretraining on gender-neutral data, finding that this reduces group disparities, i.e., promotes fairness, on VQAv2 and retrieval tasks without significantly compromising task performance., Comment: To appear in EMNLP 2024

Contrastive explanations, where one decision is explained in contrast to another, are supposed to be closer to how humans explain a decision than non-contrastive explanations, where the decision is not necessarily referenced to an alternative. This claim has never been empirically validated. We analyze four English text-classification datasets (SST2, DynaSent, BIOS and DBpedia-Animals). We fine-tune and extract explanations from three different models (RoBERTa, GTP-2, and T5), each in three different sizes and apply three post-hoc explainability methods (LRP, GradientxInput, GradNorm). We furthermore collect and release human rationale annotations for a subset of 100 samples from the BIOS dataset for contrastive and non-contrastive settings. A cross-comparison between model-based rationales and human annotations, both in contrastive and non-contrastive settings, yields a high agreement between the two settings for models as well as for humans. Moreover, model-based explanations computed in both settings align equally well with human rationales. Thus, we empirically find that humans do not necessarily explain in a contrastive manner.9 pages, long paper at ACL 2022 proceedings., Comment: 9 pages, long paper at EMNLP 2023 proceedings

We show that a quantum correlation $p$ is in a face of the nonsignaling polytope with no local points if and only if $p$ has nonlocal content 1, if and only if $p$ allows for a Greenberger-Horne-Zeilinger-like proof, and if and only if $p$ provides a perfect strategy for a nonlocal game. That is, face nonsignaling (FNS) correlations, full nonlocality (FN), all-versus nothing (AVN) proofs, and pseudotelepathy (PT) are equivalent. This shows that different resources behind a wide variety of fundamental results are in fact the same resource. We demonstrate that quantum correlations with FNS=FN=AVN=PT do not need to maximally violate a tight Bell inequality. We introduce a method for identifying quantum FNS=FN=AVN=PT correlations and use it to prove quantum mechanics does not allow for FNS=FN=AVN=PT neither in the (3,3;3,2) nor in the (3,2;3,4) Bell scenarios. This solves an open problem that, due to the FNS=FN=AVN=PT equivalence, has implications in several fields., Comment: 5+8 pages, 2+2 figures

Let $P$ be a set of at most $n$ points and let $R$ be a set of at most $n$ geometric ranges, such as for example disks or rectangles, where each $p \in P$ has an associated supply $s_{p} > 0$, and each $r \in R$ has an associated demand $d_{r} > 0$. A (many-to-many) matching is a set $\mathcal{A}$ of ordered triples $(p,r,a_{pr}) \in P \times R \times \mathbb{R}_{>0}$ such that $p \in r$ and the $a_{pr}$'s satisfy the constraints given by the supplies and demands. We show how to compute a maximum matching, that is, a matching maximizing $\sum_{(p,r,a_{pr}) \in \mathcal{A}} a_{pr}$. Using our techniques, we can also solve minimum bottleneck problems, such as computing a perfect matching between a set of $n$ red points $P$ and a set of $n$ blue points $Q$ that minimizes the length of the longest edge. For the $L_\infty$-metric, we can do this in time $O(n^{1+\varepsilon})$ in any fixed dimension, for the $L_2$-metric in the plane in time $O(n^{4/3 + \varepsilon})$, for any $\varepsilon > 0$.