Search

Differentiable simulators provide analytic gradients, enabling more sample-efficient learning algorithms and paving the way for data intensive learning tasks such as learning from images. In this work, we demonstrate that locomotion policies trained with analytic gradients from a differentiable simulator can be successfully transferred to the real world. Typically, simulators that offer informative gradients lack the physical accuracy needed for sim-to-real transfer, and vice-versa. A key factor in our success is a smooth contact model that combines informative gradients with physical accuracy, ensuring effective transfer of learned behaviors. To the best of our knowledge, this is the first time a real quadrupedal robot is able to locomote after training exclusively in a differentiable simulation., Comment: Presented at the CoRL 2024 Workshop 'Differentiable Optimization Everywhere'

Machine unlearning -- efficiently removing the effect of a small "forget set" of training data on a pre-trained machine learning model -- has recently attracted significant research interest. Despite this interest, however, recent work shows that existing machine unlearning techniques do not hold up to thorough evaluation in non-convex settings. In this work, we introduce a new machine unlearning technique that exhibits strong empirical performance even in such challenging settings. Our starting point is the perspective that the goal of unlearning is to produce a model whose outputs are statistically indistinguishable from those of a model re-trained on all but the forget set. This perspective naturally suggests a reduction from the unlearning problem to that of data attribution, where the goal is to predict the effect of changing the training set on a model's outputs. Thus motivated, we propose the following meta-algorithm, which we call Datamodel Matching (DMM): given a trained model, we (a) use data attribution to predict the output of the model if it were re-trained on all but the forget set points; then (b) fine-tune the pre-trained model to match these predicted outputs. In a simple convex setting, we show how this approach provably outperforms a variety of iterative unlearning algorithms. Empirically, we use a combination of existing evaluations and a new metric based on the KL-divergence to show that even in non-convex settings, DMM achieves strong unlearning performance relative to existing algorithms. An added benefit of DMM is that it is a meta-algorithm, in the sense that future advances in data attribution translate directly into better unlearning algorithms, pointing to a clear direction for future progress in unlearning.

The Zirconium (Z = 40) isotopic chain has attracted interest for more than four decades. The abrupt lowering of the energy of the first $2^+$ state and the increase in the transition strength B(E2; $2_1^\rightarrow 0_1^+$ going from $^{98}$Zr to $^{100}$Zr has been the first example of "quantum phase transition" in nuclear shapes, which has few equivalents in the nuclear chart. Although a multitude of experiments have been performed to measure nuclear properties related to nuclear shapes and collectivity in the region, none of the measured lifetimes were obtained using the Recoil Distance Doppler Shift method in the $\gamma\gamma$-coincidence mode where a gate on the direct feeding transition of the state of interest allows a strict control of systematical errors. This work reports the results of lifetime measurements for the first yrast excited states in $^{98-104}$Zr carried out to extract reduced transition probabilities. The new lifetime values in $\gamma\gamma$-coincidence and $\gamma$-single mode are compared with the results of former experiments. Recent predictions of the Interacting Boson Model with Configuration Mixing, the Symmetry Conserving Configuration Mixing model based on the Hartree-Fock-Bogoliubov approach and the Monte Carlo Shell Model are presented and compared with the experimental data.

Atmospheric pressure plasma jets (APPJs) are versatile devices with numerous applications. This work focuses on APPJs generated at the tip of long, flexible tubes using the jet transfer technique. The plasma source consists of a primary discharge and a secondary discharge forming the plasma jet. Discharge power measurements were carried out in a way that it was possible to separate the contribution of the primary discharge from the total power dissipated by the plasma source. Both power and effective current were analyzed under different operating conditions. The results show that the variation of the primary discharge power is much lower than the power dissipated by the plasma jet. Additionally, the electrical characteristics of the plasma device were analyzed. Notable differences were observed between the negative and positive phases of the discharge, with a more resistive load in the negative one, which suggests that the electrical equivalent circuit model changes according to the voltage polarity., Comment: 27 pages, 23 figures

We consider the two-dimensional nonlinear Schr\"odinger equation with point interaction and we establish a local well-posedness theory, including blow-up alternative and continuous dependence on the initial data in the energy space. We provide a proof by employing a Kato's method along with Hardy inequalities with logarithmic correction. Moreover, we establish finite time blow-up for solutions with positive energy and infinite variance., Comment: 22 pages, submitted

Neural Algorithmic Reasoning (NAR) research has demonstrated that graph neural networks (GNNs) could learn to execute classical algorithms. However, most previous approaches have always used a recurrent architecture, where each iteration of the GNN matches an iteration of the algorithm. In this paper we study neurally solving algorithms from a different perspective: since the algorithm's solution is often an equilibrium, it is possible to find the solution directly by solving an equilibrium equation. Our approach requires no information on the ground-truth number of steps of the algorithm, both during train and test time. Furthermore, the proposed method improves the performance of GNNs on executing algorithms and is a step towards speeding up existing NAR models. Our empirical evidence, leveraging algorithms from the CLRS-30 benchmark, validates that one can train a network to solve algorithmic problems by directly finding the equilibrium. We discuss the practical implementation of such models and propose regularisations to improve the performance of these equilibrium reasoners.

The first very long baseline interferometry (VLBI) detections at 870$\mu$m wavelength (345$\,$GHz frequency) are reported, achieving the highest diffraction-limited angular resolution yet obtained from the surface of the Earth, and the highest-frequency example of the VLBI technique to date. These include strong detections for multiple sources observed on inter-continental baselines between telescopes in Chile, Hawaii, and Spain, obtained during observations in October 2018. The longest-baseline detections approach 11$\,$G$\lambda$ corresponding to an angular resolution, or fringe spacing, of 19$\mu$as. The Allan deviation of the visibility phase at 870$\mu$m is comparable to that at 1.3$\,$mm on the relevant integration time scales between 2 and 100$\,$s. The detections confirm that the sensitivity and signal chain stability of stations in the Event Horizon Telescope (EHT) array are suitable for VLBI observations at 870$\mu$m. Operation at this short wavelength, combined with anticipated enhancements of the EHT, will lead to a unique high angular resolution instrument for black hole studies, capable of resolving the event horizons of supermassive black holes in both space and time., Comment: Corresponding author: S. Doeleman

In this paper, we study the Hamiltonian system with Hamiltonian \\ $H=\frac{1}{2}(p_r^2+p_z^2)+V_6(r,z)$, where $V_6(r,z)=r^6+Ar^2z^4+Dr^3z^3+Br^4z^2+Cz^6$, ($A,\, B,\, C,\, D \in \mathbb{R}$) for meromorphic integrability. The technique is an application of the Ziglin-Moralez-Ruiz-Ramis-Simo Theory. The difficulty that arises in solving this problem is related to the complex multivalued of the partial solution used. This difficulty is overcome by a little trick from algebraic geometry.

We consider bootstrap inference in predictive (or Granger-causality) regressions when the parameter of interest may lie on the boundary of the parameter space, here defined by means of a smooth inequality constraint. For instance, this situation occurs when the definition of the parameter space allows for the cases of either no predictability or sign-restricted predictability. We show that in this context constrained estimation gives rise to bootstrap statistics whose limit distribution is, in general, random, and thus distinct from the limit null distribution of the original statistics of interest. This is due to both (i) the possible location of the true parameter vector on the boundary of the parameter space, and (ii) the possible non-stationarity of the posited predicting (resp. Granger-causing) variable. We discuss a modification of the standard fixed-regressor wild bootstrap scheme where the bootstrap parameter space is shifted by a data-dependent function in order to eliminate the portion of limiting bootstrap randomness attributable to the boundary, and prove validity of the associated bootstrap inference under non-stationarity of the predicting variable as the only remaining source of limiting bootstrap randomness. Our approach, which is initially presented in a simple location model, has bearing on inference in parameter-on-the-boundary situations beyond the predictive regression problem.

Achieving high efficiency in modern photorealistic rendering hinges on using Monte Carlo sampling distributions that closely approximate the illumination integral estimated for every pixel. Samples are typically generated from a set of simple distributions, each targeting a different factor in the integrand, which are combined via multiple importance sampling. The resulting mixture distribution can be far from the actual product of all factors, leading to sub-optimal variance even for direct-illumination estimation. We present a learning-based method that uses normalizing flows to efficiently importance sample illumination product integrals, e.g., the product of environment lighting and material terms. Our sampler composes a flow head warp with an emitter tail warp. The small conditional head warp is represented by a neural spline flow, while the large unconditional tail is discretized per environment map and its evaluation is instant. If the conditioning is low-dimensional, the head warp can be also discretized to achieve even better performance. We demonstrate variance reduction over prior methods on a range of applications comprising complex geometry, materials and illumination., Comment: Published in ACM SIGGRAPH Asia 2024 Conference Papers. Project page: https://joeylitalien.github.io/publications/warp

Neural Algorithmic Reasoning (NAR) aims to optimize classical algorithms. However, canonical implementations of NAR train neural networks to return only a single solution, even when there are multiple correct solutions to a problem, such as single-source shortest paths. For some applications, it is desirable to recover more than one correct solution. To that end, we give the first method for NAR with multiple solutions. We demonstrate our method on two classical algorithms: Bellman-Ford (BF) and Depth-First Search (DFS), favouring deeper insight into two algorithms over a broader survey of algorithms. This method involves generating appropriate training data as well as sampling and validating solutions from model output. Each step of our method, which can serve as a framework for neural algorithmic reasoning beyond the tasks presented in this paper, might be of independent interest to the field and our results represent the first attempt at this task in the NAR literature.

How do language models use information provided as context when generating a response? Can we infer whether a particular generated statement is actually grounded in the context, a misinterpretation, or fabricated? To help answer these questions, we introduce the problem of context attribution: pinpointing the parts of the context (if any) that led a model to generate a particular statement. We then present ContextCite, a simple and scalable method for context attribution that can be applied on top of any existing language model. Finally, we showcase the utility of ContextCite through three applications: (1) helping verify generated statements (2) improving response quality by pruning the context and (3) detecting poisoning attacks. We provide code for ContextCite at https://github.com/MadryLab/context-cite.

Self-organization in complex systems is a process in which randomness is reduced and emergent structures appear that allow the system to function in a more competitive way with other states of the system or with other systems. It occurs only in the presence of energy gradients, facilitating energy transmission through the system and entropy production. Being a dynamic process, self-organization requires a dynamic measure and dynamic principles. The principles of decreasing unit action and increasing total action are two dynamic variational principles that are viable to utilize in a self-organizing system. Based on this, average action efficiency can serve as a quantitative measure of the degree of self-organization. Positive feedback loops connect this measure with all other characteristics of a complex system, providing all of them with a mechanism for exponential growth, and indicating power law relationships between each of them as confirmed by data and simulations. In this study, we apply those principles and the model to agent-based simulations. We find that those principles explain self-organization well and that the results confirm the model. By measuring action efficiency we can have a new answer to the question: "What is complexity and how complex is a system?". This work shows the explanatory and predictive power of those models, which can help understand and design better complex systems.

Glossy objects present a significant challenge for 3D reconstruction from multi-view input images under natural lighting. In this paper, we introduce PBIR-NIE, an inverse rendering framework designed to holistically capture the geometry, material attributes, and surrounding illumination of such objects. We propose a novel parallax-aware non-distant environment map as a lightweight and efficient lighting representation, accurately modeling the near-field background of the scene, which is commonly encountered in real-world capture setups. This feature allows our framework to accommodate complex parallax effects beyond the capabilities of standard infinite-distance environment maps. Our method optimizes an underlying signed distance field (SDF) through physics-based differentiable rendering, seamlessly connecting surface gradients between a triangle mesh and the SDF via neural implicit evolution (NIE). To address the intricacies of highly glossy BRDFs in differentiable rendering, we integrate the antithetic sampling algorithm to mitigate variance in the Monte Carlo gradient estimator. Consequently, our framework exhibits robust capabilities in handling glossy object reconstruction, showcasing superior quality in geometry, relighting, and material estimation.

The increased use of large language models (LLMs) across a variety of real-world applications calls for automatic tools to check the factual accuracy of their outputs, as LLMs often hallucinate. This is difficult as it requires assessing the factuality of free-form open-domain responses. While there has been a lot of research on this topic, different papers use different evaluation benchmarks and measures, which makes them hard to compare and hampers future progress. To mitigate these issues, we developed OpenFactCheck, a unified framework, with three modules: (i) RESPONSEEVAL, which allows users to easily customize an automatic fact-checking system and to assess the factuality of all claims in an input document using that system, (ii) LLMEVAL, which assesses the overall factuality of an LLM, and (iii) CHECKEREVAL, a module to evaluate automatic fact-checking systems. OpenFactCheck is open-sourced (https://github.com/mbzuai-nlp/openfactcheck) and publicly released as a Python library (https://pypi.org/project/openfactcheck/) and also as a web service (http://app.openfactcheck.com). A video describing the system is available at https://youtu.be/-i9VKL0HleI., Comment: 11 pages, 4 Figures, 3 Tables, Accepted at EMNLP 2024 System Demonstration. arXiv admin note: substantial text overlap with arXiv:2405.05583

In this work, we introduce Gemma 2, a new addition to the Gemma family of lightweight, state-of-the-art open models, ranging in scale from 2 billion to 27 billion parameters. In this new version, we apply several known technical modifications to the Transformer architecture, such as interleaving local-global attentions (Beltagy et al., 2020a) and group-query attention (Ainslie et al., 2023). We also train the 2B and 9B models with knowledge distillation (Hinton et al., 2015) instead of next token prediction. The resulting models deliver the best performance for their size, and even offer competitive alternatives to models that are 2-3 times bigger. We release all our models to the community.

During the last decade and a half, the new generation spectropolarimeter Narval at Pic du Midi, France allowed the study of weak magnetic fields in cool giant stars that are fairly evolved after main sequence. We present a short summary on the recent knowledge on the magnetic fields and activity in giants situated in the upper right part of the Hertzsprung-Russel (H-R) diagram and discuss on the possible mechanisms for magnetic field generation in the asymptotic giants branch (AGB) and post-AGB stars., Comment: Accepted for publication in Bulgarian Astronomical Journal, volume 42 (2025)

We introduce a theoretical and practical framework for efficient importance sampling of mini-batch samples for gradient estimation from single and multiple probability distributions. To handle noisy gradients, our framework dynamically evolves the importance distribution during training by utilizing a self-adaptive metric. Our framework combines multiple, diverse sampling distributions, each tailored to specific parameter gradients. This approach facilitates the importance sampling of vector-valued gradient estimation. Rather than naively combining multiple distributions, our framework involves optimally weighting data contribution across multiple distributions. This adapted combination of multiple importance yields superior gradient estimates, leading to faster training convergence. We demonstrate the effectiveness of our approach through empirical evaluations across a range of optimization tasks like classification and regression on both image and point cloud datasets., Comment: 13 pages, 11 figures

Reinforcement Learning (RL) has achieved impressive results on complex tasks but struggles in multi-task settings with different embodiments. World models offer scalability by learning a simulation of the environment, yet they often rely on inefficient gradient-free optimization methods. We introduce Policy learning with large World Models (PWM), a novel model-based RL algorithm that learns continuous control policies from large multi-task world models. By pre-training the world model on offline data and using it for first-order gradient policy learning, PWM effectively solves tasks with up to 152 action dimensions and outperforms methods using ground-truth dynamics. Additionally, PWM scales to an 80-task setting, achieving up to 27% higher rewards than existing baselines without the need for expensive online planning. Visualizations and code available at https://www.imgeorgiev.com/pwm, Comment: Visualizations and code available at https://www.imgeorgiev.com/pwm

Machine learning models can fail on subgroups that are underrepresented during training. While techniques such as dataset balancing can improve performance on underperforming groups, they require access to training group annotations and can end up removing large portions of the dataset. In this paper, we introduce Data Debiasing with Datamodels (D3M), a debiasing approach which isolates and removes specific training examples that drive the model's failures on minority groups. Our approach enables us to efficiently train debiased classifiers while removing only a small number of examples, and does not require training group annotations or additional hyperparameter tuning.

Conventional rendering techniques are primarily designed and optimized for single-frame rendering. In practical applications, such as scene editing and animation rendering, users frequently encounter scenes where only a small portion is modified between consecutive frames. In this paper, we develop a novel approach to incremental re-rendering of scenes with dynamic objects, where only a small part of a scene moves from one frame to the next. We formulate the difference (or residual) in the image between two frames as a (correlated) light-transport integral which we call the residual path integral. Efficient numerical solution of this integral then involves (1)~devising importance sampling strategies to focus on paths with non-zero residual-transport contributions and (2)~choosing appropriate mappings between the native path spaces of the two frames. We introduce a set of path importance sampling strategies that trace from the moving object(s) which are the sources of residual energy. We explore path mapping strategies that generalize those from gradient-domain path tracing to our importance sampling techniques specially for dynamic scenes. Additionally, our formulation can be applied to material editing as a simpler special case. We demonstrate speed-ups over previous correlated sampling of path differences and over rendering the new frame independently. Our formulation brings new insights into the re-rendering problem and paves the way for devising new types of sampling techniques and path mappings with different trade-offs., Comment: 14 pages, 13 figures

In this paper, we consider solutions to the following fourth order anisotropic nonlinear Schr\"odinger equation in $\R \times \R^2$, $$ \left\{ \begin{aligned} &\textnormal{i}\partial_t\psi+\partial_{xx} \psi-\partial_{yyyy} \psi +|\psi|^{p-2} \psi=0, \\ &\psi(0)=\psi_0 \in H^{1,2}(\R^2), \end{aligned} \right. $$ where $p>2$. First we prove the local/global well-posedness and blowup of solutions to the Cauchy problem for the anisotropic nonlinear Schr\"odinger equation. Then we establish the existence, axial symmetry, exponential decay and orbital stability/instability of standing waves to the anisotropic nonlinear Schr\"odinger equation. The pictures are considerably different from the ones for the isotropic nonlinear Schr\"odinger equations. The results are easily extendable to the higher dimensional case., Comment: 22 pages

Ge1-xSnx and Si1-x-yGeySnx alloys are promising materials for future opto- and nanoelectronics applications. These alloys enable effective band-gap engineering, broad adjustability of their lattice parameter, exhibit much higher carrier mobility than pure Si, and are compatible with the CMOS technology. Unfortunately, the equilibrium solid solubility of Sn in Si1-xGex is less than 1% and the pseudomorphic growth of Si1-x-yGeySnx on Ge or Si can cause in-plane compressive strain in the grown layer, degrading the superior properties of these alloys. Therefore, post-growth strain engineering by ultrafast non-equilibrium thermal treatments like pulse laser annealing (PLA) is needed to improve the layer quality. In this article, Ge0.94Sn0.06 and Si0.14Ge0.8Sn0.06 thin films grown on silicon-on-insulator substrates by molecular beam epitaxy were post growth thermally treated by PLA. The material is analyzed before and after the thermal treatments by transmission electron microscopy, X-ray diffraction (XRD), Rutherford backscattering spectrometry, secondary ion mass spectrometry, and Hall effect measurements. It is shown that after annealing, the material is single-crystalline with improved crystallinity than the as-grown layer. This is reflected in a significantly increased XRD reflection intensity, well-ordered atomic pillars, and increased active carrier concentrations up to 4x1019 cm-3.

For many years, Si1-yGey alloys have been applied in the semiconductor industry due to the ability to adjust the performance of Si-based nanoelectronic devices. Following this alloying approach of group-IV semiconductors, adding tin (Sn) into the alloy appears as the obvious next step, which leads to additional possibilities for tailoring the material properties. Adding Sn enables effective band gap and strain engineering and can improve the carrier mobilities, which makes Si1-x-yGeySnx alloys promising candidates for future opto- and nanoelectronics applications. The bottom-up approach for epitaxial growth of Si1-x-yGeySnx, e.g., by chemical vapor deposition and molecular beam epitaxy, allows tuning the material properties in the growth direction only; the realization of local material modifications to generate lateral heterostructures with such a bottom-up approach is extremely elaborate, since it would require the use of lithography, etching, and either selective epitaxy or epitaxy and chemical-mechanical polishing giving rise to interface issues, non-planar substrates, etc. This article shows the possibility of fabricating Si1-x-yGeySnx alloys by Sn ion beam implantation into Si1-yGey layers followed by millisecond-range flash lamp annealing (FLA). The materials are investigated by Rutherford backscattering spectrometry, micro Raman spectroscopy, X-ray diffraction, and transmission electron microscopy. The fabrication approach was adapted to ultra-thin Si1-yGey layers on silicon-on-insulator substrates. The results show the fabrication of single-crystalline Si1-x-yGeySnx with up to 2.3 at.% incorporated Sn without any indication of Sn segregation after recrystallization via FLA. Finally, we exhibit the possibility of implanting Sn locally in ultra-thin Si1-yGey films by masking unstructured regions on the chip.

Do neural networks learn to implement algorithms such as look-ahead or search "in the wild"? Or do they rely purely on collections of simple heuristics? We present evidence of learned look-ahead in the policy network of Leela Chess Zero, the currently strongest neural chess engine. We find that Leela internally represents future optimal moves and that these representations are crucial for its final output in certain board states. Concretely, we exploit the fact that Leela is a transformer that treats every chessboard square like a token in language models, and give three lines of evidence (1) activations on certain squares of future moves are unusually important causally; (2) we find attention heads that move important information "forward and backward in time," e.g., from squares of future moves to squares of earlier ones; and (3) we train a simple probe that can predict the optimal move 2 turns ahead with 92% accuracy (in board states where Leela finds a single best line). These findings are an existence proof of learned look-ahead in neural networks and might be a step towards a better understanding of their capabilities., Comment: Project page: https://leela-interp.github.io/

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

Advanced experiments using THz pump and X-ray probe pulses at modern free-electron lasers (FELs) like the European X-ray FEL require a frequency-tunable, high-power, narrow-band THz source maintaining the repetition rate and pulse structure of the X-ray pulses. This paper reports the first results from a THz source, that is based on a single-pass high-gain THz FEL operating with a central wavelength of 100 micrometers. The THz FEL prototype is currently in operation at the Photo Injector Test facility at DESY in Zeuthen (PITZ) and uses the same type of electron source as the European XFEL photo injector. A self-amplified spontaneous emission (SASE) FEL was envisioned as the main mechanism for generating the THz pulses. Although the THz FEL at PITZ is supposed to use the same mechanism as at X-ray facilities, it cannot be considered as a simple scaling of the radiation wavelength because there is a large difference in the number of electrons per radiation wavelength, which is five orders of magnitude higher for the THz case. The bunching factor arising from the electron beam current profile contributes strongly to the initial spontaneous emission starting the FEL process. Proof-of-principle experiments were done at PITZ using an LCLS-I undulator to generate the first high-power, high-repetition-rate single-pass THz FEL radiation. Electron bunches with a beam energy of ~17 MeV and a bunch charge of up to several nC are used to generate THz pulses with a pulse energy of several tens of microjoules. For example, for an electron beam with a charge of ~2.4 nC, more than 100 microjoules were generated at a central wavelength of 100 micrometers. The narrowband spectrum was also demonstrated by spectral measurements. These proof-of-principle experiments pave the way for a tunable, high-repetition-rate THz source providing pulses with energies in the millijoule range., Comment: (unpublished) for submission to PRAB

Model-Free Reinforcement Learning (MFRL), leveraging the policy gradient theorem, has demonstrated considerable success in continuous control tasks. However, these approaches are plagued by high gradient variance due to zeroth-order gradient estimation, resulting in suboptimal policies. Conversely, First-Order Model-Based Reinforcement Learning (FO-MBRL) methods employing differentiable simulation provide gradients with reduced variance but are susceptible to sampling error in scenarios involving stiff dynamics, such as physical contact. This paper investigates the source of this error and introduces Adaptive Horizon Actor-Critic (AHAC), an FO-MBRL algorithm that reduces gradient error by adapting the model-based horizon to avoid stiff dynamics. Empirical findings reveal that AHAC outperforms MFRL baselines, attaining 40% more reward across a set of locomotion tasks and efficiently scaling to high-dimensional control environments with improved wall-clock-time efficiency., Comment: Website https://adaptive-horizon-actor-critic.github.io/

Neural representations have shown spectacular ability to compress complex signals in a fraction of the raw data size. In 3D computer graphics, the bulk of a scene's memory usage is due to polygons and textures, making them ideal candidates for neural compression. Here, the main challenge lies in finding good trade-offs between efficient compression and cheap inference while minimizing training time. In the context of rendering, we adopt a ray-centric approach to this problem and devise N-BVH, a neural compression architecture designed to answer arbitrary ray queries in 3D. Our compact model is learned from the input geometry and substituted for it whenever a ray intersection is queried by a path-tracing engine. While prior neural compression methods have focused on point queries, ours proposes neural ray queries that integrate seamlessly into standard ray-tracing pipelines. At the core of our method, we employ an adaptive BVH-driven probing scheme to optimize the parameters of a multi-resolution hash grid, focusing its neural capacity on the sparse 3D occupancy swept by the original surfaces. As a result, our N-BVH can serve accurate ray queries from a representation that is more than an order of magnitude more compact, providing faithful approximations of visibility, depth, and appearance attributes. The flexibility of our method allows us to combine and overlap neural and non-neural entities within the same 3D scene and extends to appearance level of detail., Comment: 10 pages

Novel-view synthesis of specular objects like shiny metals or glossy paints remains a significant challenge. Not only the glossy appearance but also global illumination effects, including reflections of other objects in the environment, are critical components to faithfully reproduce a scene. In this paper, we present Neural Directional Encoding (NDE), a view-dependent appearance encoding of neural radiance fields (NeRF) for rendering specular objects. NDE transfers the concept of feature-grid-based spatial encoding to the angular domain, significantly improving the ability to model high-frequency angular signals. In contrast to previous methods that use encoding functions with only angular input, we additionally cone-trace spatial features to obtain a spatially varying directional encoding, which addresses the challenging interreflection effects. Extensive experiments on both synthetic and real datasets show that a NeRF model with NDE (1) outperforms the state of the art on view synthesis of specular objects, and (2) works with small networks to allow fast (real-time) inference. The project webpage and source code are available at: \url{https://lwwu2.github.io/nde/}., Comment: Accepted to CVPR 2024

This paper presents a detailed characterization of the positron beam delivered by the Beam Test Facility at Laboratori Nazionali of Frascati to the PADME experiment during Run III, which took place from October to December 2022. It showcases the methodology used to measure the main beam parameters such as the position in space, the absolute momentum scale, the beam energy spread, and its intensity through a combination of data analysis and Monte Carlo simulations. The results achieved include an absolute precision in the momentum of the beam to within $\sim$ 1-2 MeV$/c$, a relative beam energy spread below 0.25\%, and an absolute precision in the intensity of the beam at the level of 2\% percent.

The increased use of large language models (LLMs) across a variety of real-world applications calls for mechanisms to verify the factual accuracy of their outputs. Difficulties lie in assessing the factuality of free-form responses in open domains. Also, different papers use disparate evaluation benchmarks and measurements, which renders them hard to compare and hampers future progress. To mitigate these issues, we propose OpenFactCheck, a unified factuality evaluation framework for LLMs. OpenFactCheck consists of three modules: (i) CUSTCHECKER allows users to easily customize an automatic fact-checker and verify the factual correctness of documents and claims, (ii) LLMEVAL, a unified evaluation framework assesses LLM's factuality ability from various perspectives fairly, and (iii) CHECKEREVAL is an extensible solution for gauging the reliability of automatic fact-checkers' verification results using human-annotated datasets. OpenFactCheck is publicly released at https://github.com/yuxiaw/OpenFactCheck., Comment: 19 pages, 8 tables, 8 figures

The final stages of cosmic reionisation (EndEoR) are expected to be strongly regulated by the residual neutral hydrogen in the already ionised regions of the Universe. Its presence limits the mean distance that ionising photons can travel and hence, the extent of the regions that sources of ionising photons can affect. The structures containing most of this residual neutral hydrogen are typically unresolved in large-scale simulations of reionisation. Here, we investigate and compare a range of approaches for including the effect of these small-scale absorbers, also known as Lyman limit systems (LLS), in such simulations. We evaluate the impact of these different approaches on the reionisation history, the evolution of the ultraviolet background, and its fluctuations. We also compare to observational results on the distribution of Lyman-$\alpha$ opacity towards the EndEoR and the measured mean free path of ionising photons. We further consider their effect on the 21-cm power spectrum. We find that although each of the different approaches can match some of the observed probes of the final stages of reionisation, only the use of a redshift-dependent and position-dependent LLS model is able to reproduce all of them. We therefore recommend that large-scale reionisation simulations, which aim to describe both the state of the ionised and neutral intergalactic medium use such an approach, although the other, simpler approaches are applicable depending on the science goal of the simulation.

The three areas of realistic forward rendering, per-pixel inverse rendering, and generative image synthesis may seem like separate and unrelated sub-fields of graphics and vision. However, recent work has demonstrated improved estimation of per-pixel intrinsic channels (albedo, roughness, metallicity) based on a diffusion architecture; we call this the RGB$\rightarrow$X problem. We further show that the reverse problem of synthesizing realistic images given intrinsic channels, X$\rightarrow$RGB, can also be addressed in a diffusion framework. Focusing on the image domain of interior scenes, we introduce an improved diffusion model for RGB$\rightarrow$X, which also estimates lighting, as well as the first diffusion X$\rightarrow$RGB model capable of synthesizing realistic images from (full or partial) intrinsic channels. Our X$\rightarrow$RGB model explores a middle ground between traditional rendering and generative models: we can specify only certain appearance properties that should be followed, and give freedom to the model to hallucinate a plausible version of the rest. This flexibility makes it possible to use a mix of heterogeneous training datasets, which differ in the available channels. We use multiple existing datasets and extend them with our own synthetic and real data, resulting in a model capable of extracting scene properties better than previous work and of generating highly realistic images of interior scenes.

We investigate a promising conformal field theory realization scheme for topological quantum computation based on the Fibonacci anyons, which are believed to be realized as quasiparticle excitations in the $\mathbb{Z}_3$ parafermion fractional quantum Hall state in the second Landau level with filling factor $\nu=12/5$. These anyons are non-Abelian and are known to be capable of universal topological quantum computation. The quantum information is encoded in the fusion channels of pairs of such non-Abelian anyons and is protected from noise and decoherence by the topological properties of these systems.The quantum gates are realized by braiding of these anyons. We propose here an implementation of the $n$-qubit topological quantum register in terms of $2n+2$ Fibonacci anyons. The matrices emerging from the anyon exchanges, i.e. the generators of the braid group for one qubit are derived from the coordinate wave functions of a large number of electron holes and 4 Fibonacci anyons which can furthermore be represented as correlation functions in $\mathbb{Z}_3$ parafermionic two-dimensional conformal field theory. The representations of the braid groups for more than 4 anyons are obtained by fusing pairs of anyons before braiding, thus reducing eventually the system to 4 anyons., Comment: 37 pages, 10 figures, 2 tables

Fibonacci anyons provide the simplest possible model of non-Abelian fusion rules: [1] x [1] = [0] + [1]. We propose a conformal field theory construction of topological quantum registers based on Fibonacci anyons realized as quasiparticle excitations in the Z_3 parafermion fractional quantum Hall state. To this end, the results of Ardonne and Schoutens for the correlation function of n = 4 Fibonacci fields are extended to the case of arbitrary n (and 3 r electrons). Special attention is paid to the braiding properties of the obtained correlators. We explain in details the construction of a monodromy representation of the Artin braid group acting on n-point conformal blocks of Fibonacci anyons. For low n (up to n = 8), the matrices of braid group generators are displayed explicitly. A simple recursion formula makes it possible to extend without efforts the construction to any n. Finally, we construct N qubit computational spaces in terms of conformal blocks of 2N + 2 Fibonacci anyons., Comment: minor revisions (39 pages, 3 figures)

In the pursuit of room temperature quantum hardware, our study introduces a gate voltage tunable quantum wire within a tri-gated n-type junctionless MOSFET. The application of gate voltage alters the parabolic potential well of the tri-gated junctionless MOSFET, enabling modification of the nanowire's potential well profile. In the presence of light, photogenerated electrons accumulate at the center of the junctionless nanowire, aligning with the modified potential well profile influenced by gate bias. These carriers at the center are far from interfaces and experience less interfacial noise. Therefore, such clean photo-doping shows clear, repeatable peaks in current for specific gate biases compared to the dark condition, considering different operating drain-to-source voltages at room temperature. We propose that photodoping-induced subband occupation of gate tunable potential well of the nanowire is the underlying phenomenon responsible for this kind of observation. This study reveals experimental findings demonstrating gate-induced switching from semi-classical to the quantum domain, followed by the optical occupancy of electronic sub-bands at room temperature. We developed a compact model based on the Nonequilibrium Green's function formalism to understand this phenomenon in our illuminated device better. This work reveals the survival of the quantum confinement effect at room temperature in such semi-classical transport., Comment: 12 pages, 6 figures

Raman spectroscopy is widely used across scientific domains to characterize the chemical composition of samples in a non-destructive, label-free manner. Many applications entail the unmixing of signals from mixtures of molecular species to identify the individual components present and their proportions, yet conventional methods for chemometrics often struggle with complex mixture scenarios encountered in practice. Here, we develop hyperspectral unmixing algorithms based on autoencoder neural networks, and we systematically validate them using both synthetic and experimental benchmark datasets created in-house. Our results demonstrate that unmixing autoencoders provide improved accuracy, robustness and efficiency compared to standard unmixing methods. We also showcase the applicability of autoencoders to complex biological settings by showing improved biochemical characterization of volumetric Raman imaging data from a monocytic cell.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

We study the existence of positive solutions of a particular elliptic system in $\mathbb{R}^3$ composed of two coupled non linear stationary Schr\"odinger equations (NLSEs), that is $-\epsilon^2 \Delta u + V(x) u= h_v(u,v), - \epsilon^2 \Delta v + V(x) v=h_u (u,v)$. Under certain hypotheses on the potential $V$ and the non linearity $h$, we manage to prove that there exists a solution $(u_\epsilon,v_\epsilon)$ that decays exponentially with respect to local minima points of the potential and whose energy tends to concentrate around these points, as $\epsilon \to 0$. We also estimate this energy in terms of particular ground state energies. This work follows closely what is done in https://doi.org/10.1007/s00526-007-0103-z , although here we consider a more general non linearity and we restrict ourselves to the case where the domain is $\mathbb{R}^3$., Comment: This article may be downloaded for personal use only. Any other use requires prior permission of the author and AIP Publishing. This article appeared in AIP Conf.Proc. 5 April 2022; 2459 (1):030003 and may be found at https://doi.org/10.1063/5.0084041

Recent work on neural algorithmic reasoning has demonstrated that graph neural networks (GNNs) could learn to execute classical algorithms. Doing so, however, has always used a recurrent architecture, where each iteration of the GNN aligns with an algorithm's iteration. Since an algorithm's solution is often an equilibrium, we conjecture and empirically validate that one can train a network to solve algorithmic problems by directly finding the equilibrium. Note that this does not require matching each GNN iteration with a step of the algorithm.

Large language models (LLMs), especially when instruction-tuned for chat, have become part of our daily lives, freeing people from the process of searching, extracting, and integrating information from multiple sources by offering a straightforward answer to a variety of questions in a single place. Unfortunately, in many cases, LLM responses are factually incorrect, which limits their applicability in real-world scenarios. As a result, research on evaluating and improving the factuality of LLMs has attracted a lot of attention recently. In this survey, we critically analyze existing work with the aim to identify the major challenges and their associated causes, pointing out to potential solutions for improving the factuality of LLMs, and analyzing the obstacles to automated factuality evaluation for open-ended text generation. We further offer an outlook on where future research should go., Comment: 11 pages, 1 figure and 2 tables

3C84 is a nearby radio source with a complex total intensity structure, showing linear polarisation and spectral patterns. A detailed investigation of the central engine region necessitates the use of VLBI above the hitherto available maximum frequency of 86GHz. Using ultrahigh resolution VLBI observations at the highest available frequency of 228GHz, we aim to directly detect compact structures and understand the physical conditions in the compact region of 3C84. We used EHT 228GHz observations and, given the limited (u,v)-coverage, applied geometric model fitting to the data. We also employed quasi-simultaneously observed, multi-frequency VLBI data for the source in order to carry out a comprehensive analysis of the core structure. We report the detection of a highly ordered, strong magnetic field around the central, SMBH of 3C84. The brightness temperature analysis suggests that the system is in equipartition. We determined a turnover frequency of $\nu_m=(113\pm4)$GHz, a corresponding synchrotron self-absorbed magnetic field of $B_{SSA}=(2.9\pm1.6)$G, and an equipartition magnetic field of $B_{eq}=(5.2\pm0.6)$G. Three components are resolved with the highest fractional polarisation detected for this object ($m_\textrm{net}=(17.0\pm3.9)$%). The positions of the components are compatible with those seen in low-frequency VLBI observations since 2017-2018. We report a steeply negative slope of the spectrum at 228GHz. We used these findings to test models of jet formation, propagation, and Faraday rotation in 3C84. The findings of our investigation into different flow geometries and black hole spins support an advection-dominated accretion flow in a magnetically arrested state around a rapidly rotating supermassive black hole as a model of the jet-launching system in the core of 3C84. However, systematic uncertainties due to the limited (u,v)-coverage, however, cannot be ignored., Comment: 15 pages, 6 figures, published in A&A

The work deals with the Ericksen-Leslie model for nematic liquid crystals on the whole space, the half-space and on exterior domains with smooth boundary. The crystal orientation is described by a unit vector that is a small perturbation of a fixed constant vector. We prove, through a combination of energy method with dispersive a priori estimates, a local existence and a global existence for small initial data by a contraction argument asking low regularity assumptions on the initial condition., Comment: 38 pages

We apply the inverse Gertsenshtein effect, i.e., the graviton-photon conversion in the presence of a magnetic field, to constrain high-frequency gravitational waves (HFGWs). Using existing astrophysical measurements, we compute upper limits on the GW energy densities $\Omega_{\rm GW}$ at 16 different frequency bands. Given the observed magnetisation of galaxy clusters with field strength $B\sim\mu{\rm G}$ correlated on $\mathcal{O}(10)\,{\rm kpc}$ scales, we estimate HFGW constraints in the $\mathcal{O}(10^2)\,{\rm GHz}$ regime to be $\Omega_{\rm GW}\lesssim10^{16}$ with the temperature measurements of the Atacama Cosmology Telescope (ACT). Similarly, we conservatively obtain $\Omega_{\rm GW}\lesssim10^{13} (10^{11})$ in the $\mathcal{O}(10^2)\,{\rm MHz}$ ($\mathcal{O}(10)\,{\rm GHz}$) regime by assuming uniform magnetic field with strength $B\sim0.1\,{\rm nG}$ and saturating the excess signal over the Cosmic Microwave Background (CMB) reported by radio telescopes such as the Experiment to Detect the Global EoR Signature (EDGES), LOw Frequency ARray (LOFAR), and Murchison Widefield Array (MWA), and the balloon-borne second generation Absolute Radiometer for Cosmology, Astrophysics, and Diffuse Emission (ARCADE2) with graviton-induced photons. The upcoming Square Kilometer Array (SKA) can tighten these constraints by roughly 10 orders of magnitude, which will be a step closer to reaching the critical value of $\Omega_{\rm GW} = 1$ or the Big Bang Nucleosynthesis (BBN) bound of $\Omega_{\rm GW}\simeq1.2\times10^{-6}$. We point to future improvement of the SKA forecast and estimate that proposed CMB measurement at the level of $\mathcal{O}(10^{0-2})\,{\rm nK}$, such as Primordial Inflation Explorer (PIXIE) and Voyage 2050, are needed to viably detect stochastic backgrounds of HFGWs., Comment: 31 pages, 4 figures, 3 tables

We present a detailed long-term study of the single M6 III giant RZ~Ari to obtain direct and simultaneous measurements of the magnetic field, activity indicators, and radial velocity in order to infer the origin of its activity. We study its magnetic activity in the context of stellar evolution, and for this purpose, we also refined its evolutionary status and Li abundance. In general, for the M giants, little is known about the properties of the magnetic activity and its causes. RZ~Ari posses the strongest surface magnetic field of the known Zeeman-detected M giants and is bright enough to allow a deep study of its surface magnetic structure. The results are expected to shed light on the activity mechanism in these stars.

We analyzed TESS photometric data of the flickering-active cataclysmic star AQ Men in 2018--2019. We processed 7 sectors with 14 light curves (LCs) inside them, with a time resolution of 2 min. Aiming to study the "macro-flickering", with quasi periods (QPs) between 10 and 100 hours, we processed LCs after 55 time reduced, with a time resolution of 1.83 hours. The method, developed earlier by us, includes comparing the LCs by their statistical and fractal parameters, as well as revealing QPs by minima of structure functions and relevant maxima of autocorrelation functions. We distinguish the known high state of AQ Men, in the sectors ## 01, 05, 08, 12, 13, as well as the low state, in sectors #19, #20. In the low state the LCs show noticeable eclipses with period 3.4 h, lower average fluxes, higher scatters, and additional QPs. By its statistical and fractal parameters, the macro-flickering of AQ Men in the high state is similar to the ordinary flickering of 3 symbiotic binaries, studied by us earlier (see below). We found 92 QPs in the range of 20--70 h. We reveal 3 QP modes, at 20.9 h, 32.5 h and 54.1 h (1.149, 0.738 and 0.434 c/d; Fig. 7) within a standard error about 10%. The last mode is the most populated one and seems to be a manifestation of the superorbital period. Other 4 QP modes of AQ Men are added from the literature. The regularity of these 7 QP modes follows a power function with a base 1.57 and standard deviation 6.4% (Fig. 8). This power model prognosticates 5 other QP modes: 3 internal and 2 external (Table 2). The bases of the power regularity models for the flickering of the symbiotic binaries RS Oph, T CrB, and MWC 560 (however in the time scale of minutes) are 1.55, 2.0, and 1.34, respectively (Table 1). For unknown reasons in these 4 cases we find (i) regularities with (ii) different bases.

Diffusion models trained on large datasets can synthesize photo-realistic images of remarkable quality and diversity. However, attributing these images back to the training data-that is, identifying specific training examples which caused an image to be generated-remains a challenge. In this paper, we propose a framework that: (i) provides a formal notion of data attribution in the context of diffusion models, and (ii) allows us to counterfactually validate such attributions. Then, we provide a method for computing these attributions efficiently. Finally, we apply our method to find (and evaluate) such attributions for denoising diffusion probabilistic models trained on CIFAR-10 and latent diffusion models trained on MS COCO. We provide code at https://github.com/MadryLab/journey-TRAK ., Comment: 29 pages, 17 figures

During the Epoch of Reionization (EoR), the ultraviolet radiation from the first stars and galaxies ionised the neutral hydrogen of the intergalactic medium, which can emit radiation through its 21 cm hyperfine transition. This 21 cm signal is a direct probe of the first light sources in the early Universe. Measuring the 21 cm power spectrum is a key science goal for the future Square Kilometre Array (SKA), however, observing and interpreting it is a challenging task. Another high-potential probe of the EoR is the patchy kinetic Sunyaev-Zel'dovich effect (pkSZ), observed as a foreground to the primary cosmic microwave background temperature anisotropies on small scales. Despite recent promising measurements by ground-based telescopes, placing constraints on reionization from pkSZ observations is a non-trivial task, subject to strong model dependence. In this work, we propose to alleviate the difficulties in observing and interpreting the 21 cm and pkSZ power spectra by combining them. With a simple yet effective parametric model that establishes a formal connection between them, we are able to jointly fit mock 21 cm and pkSZ data points. We confirm that these two observables provide complementary information on reionization, leading to significantly improved constraints when combined. We demonstrate that with as few as two measurements of the 21 cm power spectrum with 100 hours of observations with the SKA, as well as a single $\ell=3000$ pkSZ data point, we can reconstruct the reionization history of the Universe and its morphology. We find that the reionization global history (morphology) is better constrained with two 21 cm measurements at different redshifts (scales). Therefore, a combined analysis of the two probes will give access to tighter constraints on cosmic reionization even in the early stages of 21 cm detections.

This study is intended to provide in-depth insights into how design thinking and creativity issues are understood and possibly evolve in the course of design discussions in a company context. For that purpose, we use the seminar transcripts of the Design Thinking Research Symposium 12 (DTRS12) dataset "Tech-centred Design Thinking: Perspectives from a Rising Asia," which are primarily concerned with how Korean companies implement design thinking and what role designers currently play. We employed a novel method of information processing based on constructed dynamic semantic networks to investigate the seminar discussions according to company representatives and company size. We compared the quantitative dynamics in two seminars: the first involved managerial representatives of four companies, and the second involved specialized designers and management of a design center of single company. On the basis of dynamic semantic networks, we quantified the changes in four semantic measures -- abstraction, polysemy, information content, and pairwise word similarity -- in chronologically reconstructed individual design-thinking processes. Statistical analyses show that design thinking in the seminar with four companies, exhibits significant differences in the dynamics of abstraction, polysemy, and information content, compared to the seminar with the design center of single company. Both the decrease in polysemy and abstraction and the increase in information content in the individual design-thinking processes in the seminar with four companies indicate that design managers are focused on more concrete design issues, with more information and less ambiguous content to the final design product. By contrast, specialized designers manifest more abstract thinking and appear to exhibit a slightly higher level of divergence in their design processes., Comment: 24 pages, 7 figures, 1 table