Search

Task-oriented conversational datasets often lack topic variability and linguistic diversity. However, with the advent of Large Language Models (LLMs) pretrained on extensive, multilingual and diverse text data, these limitations seem overcome. Nevertheless, their generalisability to different languages and domains in dialogue applications remains uncertain without benchmarking datasets. This paper presents a holistic annotation approach for emotion and conversational quality in the context of bilingual customer support conversations. By performing annotations that take into consideration the complete instances that compose a conversation, one can form a broader perspective of the dialogue as a whole. Furthermore, it provides a unique and valuable resource for the development of text classification models. To this end, we present benchmarks for Emotion Recognition and Dialogue Quality Estimation and show that further research is needed to leverage these models in a production setting., Comment: Accepted at GEM (EMNLP Workshop)

Satellite communications (SatCom) are crucial for global connectivity, especially in the era of emerging technologies like 6G and narrowing the digital divide. Traditional SatCom systems struggle with efficient resource management due to static multibeam configurations, hindering quality of service (QoS) amidst dynamic traffic demands. This paper introduces an innovative solution - real-time adaptive beamforming on multibeam satellites with software-defined payloads in geostationary orbit (GEO). Utilizing a Direct Radiating Array (DRA) with circular polarization in the 17.7 - 20.2 GHz band, the paper outlines DRA design and a supervised learning-based algorithm for on-board beamforming. This adaptive approach not only meets precise beam projection needs but also dynamically adjusts beamwidth, minimizes sidelobe levels (SLL), and optimizes effective isotropic radiated power (EIRP)., Comment: conference paper

The atmospheric dynamics of tidally-locked hot Jupiters is characterized by strong equatorial winds. Understanding the interaction between global circulation and chemistry is crucial in atmospheric studies and interpreting observations. Two-dimensional (2D) photochemical transport models shed light on how the atmospheric composition depends on circulation. In this paper, we introduce the 2D photochemical (horizontal and vertical) transport model, VULCAN 2D, which improves on the pseudo-2D approaches by allowing for non-uniform zonal winds. We extensively validate our VULCAN 2D with analytical solutions and benchmark comparisons. Applications to HD 189733 b and HD 209458 b reveal a transition in mixing regimes: horizontal transport predominates below $\sim$0.1 mbar while vertical mixing is more important at higher altitudes above 0.1 mbar. Motivated by the previously inferred carbon-rich atmosphere, we find that HD 209458 b with super-solar carbon-to-oxygen ratio (C/O) exhibits pronounced C$_2$H$_4$ absorption on the morning limb but not on the evening limb, owing to horizontal transport from the nightside. We discuss when a pseudo-2D approach is a valid assumption and its inherent limitations. Finally, we demonstrate the effect of horizontal transport in transmission observations and its impact on the morning-evening limb asymmetry with synthetic spectra, highlighting the need to consider global transport when interpreting exoplanet atmospheres., Comment: 19 pages, 20 figures, published in ApJ

We show that if a countably generated Lie algebra $H$ does not contain isomorphic copies of certain finite-dimensional nilpotent Lie algebras $A$ and $B$ (satisfying some mild conditions), then $H$ embeds into a quotient of $A \ast B$ that is at the same time hopfian and cohopfian. This is a Lie algebraic version of an embedding theorem proved by C. Miller and P. Schupp for groups. We also prove that any finitely presentable Lie algebra is the quotient of a finitely presented, centerless, residually nilpotent and SQ-universal Lie algebra of cohomological dimension at most $2$ by an ideal that can be generated by two elements as a Lie subalgebra. This is reminiscent of the Rips construction in group theory. In both results we use the theory of Gr\"obner-Shirshov bases., Comment: 14 pages

Satellite communications, essential for modern connectivity, extend access to maritime, aeronautical, and remote areas where terrestrial networks are unfeasible. Current GEO systems distribute power and bandwidth uniformly across beams using multi-beam footprints with fractional frequency reuse. However, recent research reveals the limitations of this approach in heterogeneous traffic scenarios, leading to inefficiencies. To address this, this paper presents a machine learning (ML)-based approach to Radio Resource Management (RRM). We treat the RRM task as a regression ML problem, integrating RRM objectives and constraints into the loss function that the ML algorithm aims at minimizing. Moreover, we introduce a context-aware ML metric that evaluates the ML model's performance but also considers the impact of its resource allocation decisions on the overall performance of the communication system., Comment: in review for conference

The fintech PicPay offers a wide range of financial services to its 30 million monthly active users, with more than 50 thousand items recommended in the PicPay mobile app. In this scenario, promoting specific items that are strategic to the company can be very challenging. In this work, we present a Switching Hybrid Recommender System that combines two algorithms to effectively promote items without negatively impacting the user's experience. The results of our A/B tests show an uplift of up to 3.2\% when compared to a default recommendation strategy., Comment: 4 pages, 4 figures, submitted to ACM KDD '23 2nd Workshop on End-End Customer Journey Optimization

We extend de concept of Compton scattering to the case of plasmons. This concept was originally applied to electrons in vacuum. Here, we consider electrons in a plasma, and study the scattering properties of photon-plasmon interactions. We show that a number $n$ of plasmons with frequency $ \omega \simeq \omega_p$ is scattered by an electron, for an incident photon with frequency $\omega' \geq \omega_p$, where $\omega_p$ is the plasma frequency. We describe the general case of arbitrary $n$ and assume that Compton scattering of plasmons is intrinsically a nonlinear process. Our theoretical model is based on Volkov solutions of the Klein-Gordon equation describing the state of relativistic electrons, when the spin is ignored. We derive the corresponding scattering probability, as well as the recoil formula associated with arbitrary the final electron states. This process can be relevant to intense laser plasma interactions.

Large, high-capacity models trained on diverse datasets have shown remarkable successes on efficiently tackling downstream applications. In domains from NLP to Computer Vision, this has led to a consolidation of pretrained models, with general pretrained backbones serving as a starting point for many applications. Can such a consolidation happen in robotics? Conventionally, robotic learning methods train a separate model for every application, every robot, and even every environment. Can we instead train generalist X-robot policy that can be adapted efficiently to new robots, tasks, and environments? In this paper, we provide datasets in standardized data formats and models to make it possible to explore this possibility in the context of robotic manipulation, alongside experimental results that provide an example of effective X-robot policies. We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions, demonstrating 527 skills (160266 tasks). We show that a high-capacity model trained on this data, which we call RT-X, exhibits positive transfer and improves the capabilities of multiple robots by leveraging experience from other platforms. More details can be found on the project website https://robotics-transformer-x.github.io., Comment: Project website: https://robotics-transformer-x.github.io

We study the dispersion properties of electron plasma waves, or plasmons, which can be excited in quantum plasmas in the nonlinear regime. In order to describe nonlinear electron response to finite amplitude plasmons, we apply the Volkov approach to non-relativistic electrons. For that purpose, we use the Schr\"odinger equation and describe the electron population of a quantum plasma as a mixture of quantum states. Within the kinetic framework that we are able to derive from the Volkov solutions, we discuss the role of the wave amplitude on the nonlinear plasma response. Finally, we focus on the quantum properties of nonlinear Landau damping and study the contributions of multi-plasmon absorption and emission processes., Comment: Plasmonics, quantum Landau damping, Volkov approach, nonlinear plasma waves (to appear in Phys. Rev. E)

We introduce a new notion of embeddability between Banach spaces. By studying the classical Mazur map, we show that it is strictly weaker than the notion of coarse embeddability. We use the techniques from metric cotype introduced by M. Mendel and A. Naor to prove results about cotype preservation and complete our study of embeddability between $\ell_p$ spaces. We confront our notion with nonlinear invariants introduced by N. Kalton, which are defined in terms of concentration properties for Lipschitz maps defined on countably branching Hamming or interlaced graphs. Finally, we address the problem of the embeddability into $\ell_\infty$.

This paper delves into the application of Machine Learning (ML) techniques in the realm of 5G Non-Terrestrial Networks (5G-NTN), particularly focusing on symbol detection and equalization for the Physical Broadcast Channel (PBCH). As 5G-NTN gains prominence within the 3GPP ecosystem, ML offers significant potential to enhance wireless communication performance. To investigate these possibilities, we present ML-based models trained with both synthetic and real data from a real 5G over-the-satellite testbed. Our analysis includes examining the performance of these models under various Signal-to-Noise Ratio (SNR) scenarios and evaluating their effectiveness in symbol enhancement and channel equalization tasks. The results highlight the ML performance in controlled settings and their adaptability to real-world challenges, shedding light on the potential benefits of the application of ML in 5G-NTN.

Agents that are aware of the separation between themselves and their environments can leverage this understanding to form effective representations of visual input. We propose an approach for learning such structured representations for RL algorithms, using visual knowledge of the agent, such as its shape or mask, which is often inexpensive to obtain. This is incorporated into the RL objective using a simple auxiliary loss. We show that our method, Structured Environment-Agent Representations, outperforms state-of-the-art model-free approaches over 18 different challenging visual simulation environments spanning 5 different robots. Website at https://sear-rl.github.io/, Comment: ICML 2023. Website at https://sear-rl.github.io/

Despite significant research effort in the development of automatic dialogue evaluation metrics, little thought is given to evaluating dialogues other than in English. At the same time, ensuring metrics are invariant to semantically similar responses is also an overlooked topic. In order to achieve the desired properties of robustness and multilinguality for dialogue evaluation metrics, we propose a novel framework that takes advantage of the strengths of current evaluation models with the newly-established paradigm of prompting Large Language Models (LLMs). Empirical results show our framework achieves state of the art results in terms of mean Spearman correlation scores across several benchmarks and ranks first place on both the Robust and Multilingual tasks of the DSTC11 Track 4 "Automatic Evaluation Metrics for Open-Domain Dialogue Systems", proving the evaluation capabilities of prompted LLMs., Comment: DSTC11 best paper for Track 4

The main limiting factor in the development of robust multilingual dialogue evaluation metrics is the lack of multilingual data and the limited availability of open sourced multilingual dialogue systems. In this work, we propose a workaround for this lack of data by leveraging a strong multilingual pretrained LLM and augmenting existing English dialogue data using Machine Translation. We empirically show that the naive approach of finetuning a pretrained multilingual encoder model with translated data is insufficient to outperform the strong baseline of finetuning a multilingual model with only source data. Instead, the best approach consists in the careful curation of translated data using MT Quality Estimation metrics, excluding low quality translations that hinder its performance., Comment: SIGDIAL23

We tackle the problem of learning complex, general behaviors directly in the real world. We propose an approach for robots to efficiently learn manipulation skills using only a handful of real-world interaction trajectories from many different settings. Inspired by the success of learning from large-scale datasets in the fields of computer vision and natural language, our belief is that in order to efficiently learn, a robot must be able to leverage internet-scale, human video data. Humans interact with the world in many interesting ways, which can allow a robot to not only build an understanding of useful actions and affordances but also how these actions affect the world for manipulation. Our approach builds a structured, human-centric action space grounded in visual affordances learned from human videos. Further, we train a world model on human videos and fine-tune on a small amount of robot interaction data without any task supervision. We show that this approach of affordance-space world models enables different robots to learn various manipulation skills in complex settings, in under 30 minutes of interaction. Videos can be found at https://human-world-model.github.io, Comment: RSS 2023. Website at https://human-world-model.github.io

In this paper, we study the category of modules over the Smith algebra which are free of finite rank over the unital polynomial subalgebra generated by the Cartan element $h$ and obtain families of such simple modules of arbitrary rank. In the case of rank one we obtain a full description of the isomorphism classes, a simplicity criterion, and an algorithm to produce all composition series. We show that all such modules have finite length and describe the composition factors and their multiplicity., Comment: 16 pages

We provide a characterization of when a coarse equivalence between coarse disjoint unions of expander graphs is close to a bijective coarse equivalence. We use this to show that if the uniform Roe algebras of coarse disjoint unions of expanders graphs are isomorphic, then the metric spaces must be bijectively coarsely equivalent.

We introduce GWDALI, a new Fisher-matrix, python based software that computes likelihood gradients to forecast parameter-estimation precision of arbitrary network of terrestrial gravitational wave detectors observing compact binary coalescences. The main new feature with respect to analogous software is to assess parameter uncertainties beyond Fisher-matrix approximation, using the derivative approximation for Likelihood (DALI). The software makes optional use of the LSC algorithm library LAL and the stochastic sampling algorithm Bilby, which can be used to perform Monte-Carlo sampling of exact or approximate likelihood functions. As an example we show comparison of estimated precision measurement of selected astrophysical parameters for both the actual likelihood, and for a variety of its derivative approximations, which turn out particularly useful when the Fisher matrix is not invertible., Comment: 18 pages, 11 figures