arXiv subjects

Silvia Rossi

Publications and source records attributed to Silvia Rossi.

At least 19 recordsLinked to original sources

arXiv2026

The Low-$α$ Splash Population in the Milky Way

The Milky Way in-situ halo, also known as the Splash, consists of old (age $>$ 10 Gyr), metal-rich ([Fe/H] $> -0.7$), high-$α$ stars, i.e., thick disk-like chemistry, on halo-like orbits (eccentricity $>$ 0.6). Its origin is linked to stars formed in the disk and dynamically heated by either internal or external agents. In this work, we investigate its low-$α$ counterpart, the low-$α$ Splash, motivated by recent findings of an old thin disk population. We conjecture that any mechanism capable of heating disk stars should affect both of present-day high- and low-$α$ old populations. Using data from the APOGEE DR17 spectroscopic catalog, we identify metal-rich low-$α$ stars with halo-like kinematics similar to those of the classical high-$α$ Splash. We investigate their possible heating mechanisms using the GASTRO suite of simulations, which allows us to explore the effects of star-forming clumps as well as a major merger in the proto-disk of a Milky Way analog galaxy. Our main results show that only clumpy Milky Way models are able to produce Splash populations through scattering by clumps in the early Galaxy, including the low-$α$ counterpart, whereas the model including only the merger and without an early clumpy phase fails to produce these populations. In the models, the low-$α$ Splash corresponds to a subset of the old thin disk that was dynamically heated by the same mechanism responsible for the formation of the high-$α$ Splash.

astro-ph.GA↗

arXiv2026

Why did My Robot Just Change Personality? Prompting Guidelines for a Grounded Robot Persona in LLM-Based HRI

Large language models (LLMs) are increasingly used for verbal interaction in social robots, yet prompt design in human-robot interaction (HRI) remains underspecified. As a result, robots may present hallucinated capabilities, unclear behavioural boundaries, and misleading personas. This paper develops a framework for prompt design in LLM-based robots and introduces a structured prompt template comprising eight functional components through which robot behaviour can be specified, bounded, and adapted. The framework is grounded in a review of prior LLM-based HRI work and complemented by survey and discussion data from HRI experts gathered at the Robo-Identity workshop at IEEE RO-MAN 2025 (N=27). The qualitative findings highlight limited legibility of robot personality, the need for user adaptation, and strong ethical concerns about safety, deception, and governance. Based on these findings, we present prompting guidelines accompanied by proof-of-concept template as a structured design and reporting aid for HRI research. We argue that prompt design should be treated as a socio-technical problem rather than a minor implementation detail, requiring explicit capability boundaries, transparent behavioural assumptions, and context-sensitive safeguards to support reliable and interpretable HRI.

cs.AI↗

arXiv2026

Flexible Estimation of the Heterogeneous Non-Parametric Component in a Relative Survival Cure Model

Estimating the cure fraction in a diseased population, especially in the presence of competing mortality causes, is crucial for both patients and clinicians. It offers a valuable measure for monitoring and interpreting trends in disease outcomes. When information on the cause of death is unavailable or unreliable, the Relative Survival (RS) framework is the preferred approach for estimating Net Survival, which represents survival in a hypothetical scenario where the disease of interest is the only possible cause of death. In the context of cancer, RS often reaches a plateau, indicating that a portion of diagnosed patients is cured, as they have the same risk of dying as a comparable group of healthy individuals with similar demographic characteristics. Classical RS cure models use logistic regression to estimate the fraction of cured patients. However, this functional form is somewhat arbitrary, and misspecifying it can severely distort the resulting cure indicators. Consequently, evaluations of the efficacy of cancer treatments at the population level could be inaccurate, leading to biased decision-making regarding patient care. In this paper, we address this issue by relaxing the parametric assumption and considering flexible functions of the covariates within the framework of \textit{Generalized Models} and \textit{Neural Networks}. We design an EM algorithm for these RS cure models and conduct a simulation study to compare our proposals with the classical approach. We apply our methodology to a real-world dataset from a historical Italian cancer registry. The results demonstrate that our proposed models outperform the classical approach and provide valuable insights into the survival outcomes of Italian colon cancer patients.

stat.ME↗

arXiv2026

Resisting Humanization: Ethical Front-End Design Choices in AI for Sensitive Contexts

Ethical debates in AI have primarily focused on back-end issues such as data governance, model training, and algorithmic decision-making. Less attention has been paid to the ethical significance of front-end design choices, such as the interaction and representation-based elements through which users interact with AI systems. This gap is particularly significant for Conversational User Interfaces (CUI) based on Natural Language Processing (NLP) systems, where humanizing design elements such as dialogue-based interaction, emotive language, personality modes, and anthropomorphic metaphors are increasingly prevalent. This work argues that humanization in AI front-end design is a value-driven choice that profoundly shapes users' mental models, trust calibration, and behavioral responses. Drawing on research in human-computer interaction (HCI), conversational AI, and value-sensitive design, we examine how interfaces can play a central role in misaligning user expectations, fostering misplaced trust, and subtly undermining user autonomy, especially in vulnerable contexts. To ground this analysis, we discuss two AI systems developed by Chayn, a nonprofit organization supporting survivors of gender-based violence. Chayn is extremely cautious when building AI that interacts with or impacts survivors by operationalizing their trauma-informed design principles. This Chayn case study illustrates how ethical considerations can motivate principled restraint in interface design, challenging engagement-based norms in contemporary AI products. We argue that ethical front-end AI design is a form of procedural ethics, enacted through interaction choices rather than embedded solely in system logic.

cs.AI↗

arXiv2026

S-PLUS: Beyond Spectroscopy IV. Stellar Parameters and Elemental-abundance Ratios for Six Million Stars from DR4 and First Results for the Magellanic Clouds

We combine narrow/medium-band filter photometry from the Southern Photometric Local Universe Survey (S-PLUS) DR4 with ultra broad-band filter photometry from Gaia EDR3 to derive fundamental stellar parameters ($T_{\rm eff}$, $\log g$, [Fe/H], ages) and elemental-abundance ratios ([C/Fe] and [$α$/Fe]) for 5.4 million stars in the Galaxy (4.9 million dwarfs and 0.5 million giants), as well as for over 0.7 million red giant stars in the Large and Small Magellanic Clouds (LMC and SMC). The precisions of the abundance estimates range from 0.05-0.10 dex for metallicity in the relatively metal-rich range ([Fe/H] $> -1.0$) to 0.10-0.30 dex in the metal-poor regime ([Fe/H] $<-1.0$), 0.10-0.20\,dex for [C/Fe], and 0.05 dex for [$α$/Fe]. The stellar parameters for LMC and SMC member stars are somewhat less precise than those from the S-PLUS main survey, primarily because of the effect of high reddening. The use of both metallicity- and carbon-sensitive filters provides unbiased measurements of both [Fe/H] and [C/Fe], of particular importance for very low-metallicity ([Fe/H] $< -2.0$) stars, where carbon enhancement can lead to systematically high estimates of [Fe/H] when only a single metallicity-sensitive filter is employed. Furthermore, multiple narrow-band filters enable metallicity estimates down to [Fe/H] $\sim -4.0$ with an accuracy of around 0.3 dex, exceeding the precision typically achieved by low/medium-resolution spectroscopy. This extensive photometric dataset, combined with the other three datasets in this series, will serve as a valuable legacy resource for Milky Way and Magellanic Clouds studies.

astro-ph.SR↗

arXiv2026

HEXAR: a Hierarchical Explainability Architecture for Robots

As robotic systems become increasingly complex, the need for explainable decision-making becomes critical. Existing explainability approaches in robotics typically either focus on individual modules, which can be difficult to query from the perspective of high-level behaviour, or employ monolithic approaches, which do not exploit the modularity of robotic architectures. We present HEXAR (Hierarchical EXplainability Architecture for Robots), a novel framework that provides a plug-in, hierarchical approach to generate explanations about robotic systems. HEXAR consists of specialised component explainers using diverse explanation techniques (e.g., LLM-based reasoning, causal models, feature importance, etc) tailored to specific robot modules, orchestrated by an explainer selector that chooses the most appropriate one for a given query. We implement and evaluate HEXAR on a TIAGo robot performing assistive tasks in a home environment, comparing it against end-to-end and aggregated baseline approaches across 180 scenario-query variations. We observe that HEXAR significantly outperforms baselines in root cause identification, incorrect information exclusion, and runtime, offering a promising direction for transparent autonomous systems.

cs.RO↗

arXiv2025

PerspAct: Enhancing LLM Situated Collaboration Skills through Perspective Taking and Active Vision

Recent advances in Large Language Models (LLMs) and multimodal foundation models have significantly broadened their application in robotics and collaborative systems. However, effective multi-agent interaction necessitates robust perspective-taking capabilities, enabling models to interpret both physical and epistemic viewpoints. Current training paradigms often neglect these interactive contexts, resulting in challenges when models must reason about the subjectivity of individual perspectives or navigate environments with multiple observers. This study evaluates whether explicitly incorporating diverse points of view using the ReAct framework, an approach that integrates reasoning and acting, can enhance an LLM's ability to understand and ground the demands of other agents. We extend the classic Director task by introducing active visual exploration across a suite of seven scenarios of increasing perspective-taking complexity. These scenarios are designed to challenge the agent's capacity to resolve referential ambiguity based on visual access and interaction, under varying state representations and prompting strategies, including ReAct-style reasoning. Our results demonstrate that explicit perspective cues, combined with active exploration strategies, significantly improve the model's interpretative accuracy and collaborative effectiveness. These findings highlight the potential of integrating active perception with perspective-taking mechanisms in advancing LLMs' application in robotics and multi-agent systems, setting a foundation for future research into adaptive and context-aware AI systems.

cs.RO↗

arXiv2025

Growing Perspectives: Modelling Embodied Perspective Taking and Inner Narrative Development Using Large Language Models

Language and embodied perspective taking are essential for human collaboration, yet few computational models address both simultaneously. This work investigates the PerspAct system [1], which integrates the ReAct (Reason and Act) paradigm with Large Language Models (LLMs) to simulate developmental stages of perspective taking, grounded in Selman's theory [2]. Using an extended director task, we evaluate GPT's ability to generate internal narratives aligned with specified developmental stages, and assess how these influence collaborative performance both qualitatively (action selection) and quantitatively (task efficiency). Results show that GPT reliably produces developmentally-consistent narratives before task execution but often shifts towards more advanced stages during interaction, suggesting that language exchanges help refine internal representations. Higher developmental stages generally enhance collaborative effectiveness, while earlier stages yield more variable outcomes in complex contexts. These findings highlight the potential of integrating embodied perspective taking and language in LLMs to better model developmental dynamics and stress the importance of evaluating internal speech during combined linguistic and embodied tasks.

cs.CL↗

arXiv2025

The DECam MAGIC Survey: A Wide-field Photometric Metallicity Study of the Sculptor Dwarf Spheroidal Galaxy

The metallicity distribution function and internal chemical variations of a galaxy are fundamental to understand its formation and assembly history. In this work, we analyze photometric metallicities for 3883 stars over seven half-light radii ($\rm r_h$) in the Sculptor dwarf spheroidal (Scl dSph) galaxy, using new narrow-band imaging data from the Mapping the Ancient Galaxy in CaHK (MAGIC) survey conducted with the Dark Energy Camera (DECam) at the 4-m Blanco Telescope. This work demonstrates the scientific potential of MAGIC using the Scl dSph galaxy, one of the most well-studied satellites of the Milky Way. Our sample ranges from $\rm [Fe/H] \approx - 4.0$ to $\rm [Fe/H] \approx - 0.6$, includes six new extremely metal-poor candidates ($\rm [Fe/H] \leq -3.0$), and is almost three times larger than the largest spectroscopic metallicity dataset in the Scl dSph. Our spatially unbiased sample of metallicities provides a more accurate representation of the metallicity distribution function, revealing a more metal-rich peak than observed in the most recent spectroscopic sample. It also reveals a break in the metallicity gradient, with a strong change in the slope: from $-3.26 \pm 0.18 \rm \ dex/deg$ for stars inside $\sim 1\ \rm r_h$ to $-0.55 \pm 0.26 \rm \ dex/deg$ for the outer part of the Scl dSph. Our study demonstrates that combining photometric metallicity analysis with the wide field of view of DECam offers an efficient and unbiased approach for studying the stellar populations of dwarf galaxies in the Local Group.

astro-ph.GA↗

arXiv2025

Who Sees What? Structured Thought-Action Sequences for Epistemic Reasoning in LLMs

Recent advances in large language models (LLMs) and reasoning frameworks have opened new possibilities for improving the perspective -taking capabilities of autonomous agents. However, tasks that involve active perception, collaborative reasoning, and perspective taking (understanding what another agent can see or knows) pose persistent challenges for current LLM-based systems. This study investigates the potential of structured examples derived from transformed solution graphs generated by the Fast Downward planner to improve the performance of LLM-based agents within a ReAct framework. We propose a structured solution-processing pipeline that generates three distinct categories of examples: optimal goal paths (G-type), informative node paths (E-type), and step-by-step optimal decision sequences contrasting alternative actions (L-type). These solutions are further converted into ``thought-action'' examples by prompting an LLM to explicitly articulate the reasoning behind each decision. While L-type examples slightly reduce clarification requests and overall action steps, they do not yield consistent improvements. Agents are successful in tasks requiring basic attentional filtering but struggle in scenarios that required mentalising about occluded spaces or weighing the costs of epistemic actions. These findings suggest that structured examples alone are insufficient for robust perspective-taking, underscoring the need for explicit belief tracking, cost modelling, and richer environments to enable socially grounded collaboration in LLM-based agents.

cs.AI↗

arXiv2025

Discovery of an $\rm[Fe/H] \sim -4.8$ Star in $Gaia$ XP Spectra

We report on the discovery of GDR3_526285 ($Gaia$ DR3 Source ID 5262850721755411072), a star with $\rm[Fe/H] = -4.82 \pm 0.25$ and one of the lowest metal ($\text{atomic number} > 2$) mass fractions ever found ($Z_{\rm GDR3\_526585} \lesssim 1.0 \times 10^{-6}$). We first identified it as an ultra metal-poor (UMP; $\rm[Fe/H] < -4$) red giant-branch (RGB) star candidate in the $Gaia$ BP/RP (XP) spectro-photometric catalog ($Gaia$ $G$ magnitude $\approx$15). A combination of multi-band photometry and high-resolution spectroscopic analysis under local thermodynamic equilibrium confirmed the status of GDR3_526285 as a distant ($\approx$24 kpc from the Sun) RGB star ($T_{\rm eff} = 4596\,{\rm K}$, $\log g = 0.88$) in the Milky Way's outer halo. We obtain only an upper limit for the carbon abundance of $\rm[C/H] < -4.32$, resulting in $\rm[C/Fe] < +0.50$. A correction for the evolutionary carbon depletion ($Δ\rm[C/Fe] = +0.68$) brings the nominal carbon-to-iron ratio upper limit to $\rm[C/Fe]_{\rm cor} < +1.18$. Given its extraordinarily low [C/H], GDR3_526285 likely formed from gas cooled via dust grains rather than fine structure line cooling. The kinematics of GDR3_526285 suggests that this star was either dynamically perturbed by the infall of the Magellanic system or was formerly a member of the Magellanic Clouds and was later stripped by the Milky Way. Our results showcase the potential of an all-sky search for low-metallicity targets with $Gaia$ XP and confirm that the methodology described here is an useful "treasure map" for finding additional UMP stars.

astro-ph.GA↗

arXiv2025

The DECam MAGIC Survey: Spectroscopic Follow-up of the Most Metal-Poor Stars in the Distant Milky Way Halo

In this work, we present high-resolution spectroscopic observations for six metal-poor stars with [Fe/H]<-3 (including one with [Fe/H]<-4), selected using narrow-band Ca II HK photometry from the DECam MAGIC Survey. The spectroscopic data confirms the accuracy of the photometric metallicities and allows for the determination of chemical abundances for 16 elements, from carbon to barium. The program stars have chemical abundances consistent with this metallicity range. A kinematic/dynamical analysis suggests that all program stars belong to the distant Milky Way halo population (heliocentric distances 35 < dhelio/kpc < 55), including three with high-energy orbits that might have been associated with the Magellanic system and one, J0026-5445, having parameters consistent with being a member of the Sagittarius stream. The remaining two stars show kinematics consistent with the Gaia-Sausage/Enceladus dwarf galaxy merger. J0433-5548, with [Fe/H]=-4.12, is a carbon-enhanced ultra metal-poor star, with [C/Fe]=+1.73. This star is believed to be a bona fide second-generation star, and its chemical abundance pattern was compared with yields from metal-free supernova models. Results suggest that J0433-5548 could have been formed from a gas cloud enriched by a single supernova explosion from a ~11Mo star in the early universe. The successful identification of such objects demonstrates the reliability of photometric metallicity estimates, which can be used for target selection and statistical studies of faint targets in the Milky Way and its satellite population. These discoveries illustrate the power of measuring chemical abundances of metal-poor Milky Way halo stars to learn more about early galaxy formation and evolution.

astro-ph.GA↗

arXiv2025

Early Co-formation of the Milky Way's Thin and Thick Disks at Redshift z > 2

The Milky Way serves as a template for understanding the formation and evolution of late-type massive disk galaxies since we can obtain detailed chemical and kinematic information for large samples of individual stars. However, the early formation of the disk and the dichotomy between the chemical thick and thin disks remain under intense debate. Some mechanisms have been proposed to explain the formation of this dichotomy, such as the injection of metal-poor gas by a gas-rich merger such as Gaia-Sausage Enceladus (GSE), or by cosmic gas filaments, radial migration, and the presence of star-forming clumps at high redshift ($z > 2$). In this work, we combine astrometric data from the Gaia mission, chemical abundances from APOGEE and LAMOST spectroscopic surveys, and StarHorse ages to map the evolution of our Galaxy. The Bayesian isochrone-fitting code StarHorse can estimate ages for thousands of stars in the solar neighborhood, being most reliable for main sequence turnoff and sub-giants, computing distances and extinction simultaneously. From these samples, we show that (i) there is an old thin disk population ($>11$ Gyr) that indicates a period of co-formation between the thick and thin disks of the Milky Way before the GSE merger, i.e. the Galaxy itself could initiate the formation of a low-alpha disk without the need for a gas-rich merger, and (ii) this merger would have been important to stop the formation of stars in the thick disk.

astro-ph.GA↗

arXiv2024

Enhancing Robot Assistive Behaviour with Reinforcement Learning and Theory of Mind

The adaptation to users' preferences and the ability to infer and interpret humans' beliefs and intents, which is known as the Theory of Mind (ToM), are two crucial aspects for achieving effective human-robot collaboration. Despite its importance, very few studies have investigated the impact of adaptive robots with ToM abilities. In this work, we present an exploratory comparative study to investigate how social robots equipped with ToM abilities impact users' performance and perception. We design a two-layer architecture. The Q-learning agent on the first layer learns the robot's higher-level behaviour. On the second layer, a heuristic-based ToM infers the user's intended strategy and is responsible for implementing the robot's assistance, as well as providing the motivation behind its choice. We conducted a user study in a real-world setting, involving 56 participants who interacted with either an adaptive robot capable of ToM, or with a robot lacking such abilities. Our findings suggest that participants in the ToM condition performed better, accepted the robot's assistance more often, and perceived its ability to adapt, predict and recognise their intents to a higher degree. Our preliminary insights could inform future research and pave the way for designing more complex computation architectures for adaptive behaviour with ToM capabilities.

cs.RO↗

arXiv2024

Multimodal Coherent Explanation Generation of Robot Failures

The explainability of a robot's actions is crucial to its acceptance in social spaces. Explaining why a robot fails to complete a given task is particularly important for non-expert users to be aware of the robot's capabilities and limitations. So far, research on explaining robot failures has only considered generating textual explanations, even though several studies have shown the benefits of multimodal ones. However, a simple combination of multiple modalities may lead to semantic incoherence between the information across different modalities - a problem that is not well-studied. An incoherent multimodal explanation can be difficult to understand, and it may even become inconsistent with what the robot and the human observe and how they perform reasoning with the observations. Such inconsistencies may lead to wrong conclusions about the robot's capabilities. In this paper, we introduce an approach to generate coherent multimodal explanations by checking the logical coherence of explanations from different modalities, followed by refinements as required. We propose a classification approach for coherence assessment, where we evaluate if an explanation logically follows another. Our experiments suggest that fine-tuning a neural network that was pre-trained to recognize textual entailment, performs well for coherence assessment of multimodal explanations. Code & data: https://pradippramanick.github.io/coherent-explain/.

cs.RO↗

arXiv2024

Measuring Transparency in Intelligent Robots

As robots become increasingly integrated into our daily lives, the need to make them transparent has never been more critical. Yet, despite its importance in human-robot interaction, a standardized measure of robot transparency has been missing until now. This paper addresses this gap by presenting the first comprehensive scale to measure perceived transparency in robotic systems, available in English, German, and Italian languages. Our approach conceptualizes transparency as a multidimensional construct, encompassing explainability, legibility, predictability, and meta-understanding. The proposed scale was a product of a rigorous three-stage process involving 1,223 participants. Firstly, we generated the items of our scale, secondly, we conducted an exploratory factor analysis, and thirdly, a confirmatory factor analysis served to validate the factor structure of the newly developed TOROS scale. The final scale encompasses 26 items and comprises three factors: Illegibility, Explainability, and Predictability. TOROS demonstrates high cross-linguistic reliability, inter-factor correlation, model fit, internal consistency, and convergent validity across the three cross-national samples. This empirically validated tool enables the assessment of robot transparency and contributes to the theoretical understanding of this complex construct. By offering a standardized measure, we facilitate consistent and comparable research in human-robot interaction in which TOROS can serve as a benchmark.

cs.RO↗

arXiv2024

The S-PLUS Ultra-Short Survey: first data release

This paper presents the first public data release of the S-PLUS Ultra-Short Survey (USS), a photometric survey with short exposure times, covering approximately 9300 deg$^{2}$ of the Southern sky. The USS utilizes the Javalambre 12-band magnitude system, including narrow and medium-band and broad-band filters targeting prominent stellar spectral features. The primary objective of the USS is to identify bright, extremely metal-poor (EMP; [Fe/H] $\leq -3$) and ultra metal-poor (UMP; [Fe/H] $\leq -4$) stars for further analysis using medium- and high-resolution spectroscopy.}{This paper provides an overview of the survey observations, calibration method, data quality, and data products. Additionally, it presents the selection of EMP and UMP candidates.}{The data from the USS were reduced and calibrated using the same methods as presented in the S-PLUS DR2. An additional step was introduced, accounting for the offset between the observed magnitudes off the USS and the predicted magnitudes from the very low-resolution Gaia XP spectra.}{This first release contains data for 163 observed fields totaling $\sim$324 deg$^{2}$ along the Celestial Equator. The magnitudes obtained from the USS are well-calibrated, showing a difference of $\sim 15$ mmag compared to the predicted magnitudes by the GaiaXPy toolkit. By combining colors and magnitudes, 140 candidates for EMP or UMP have been identified for follow-up studies.}{The S-PLUS USS DR1 is an important milestone in the search for bright metal-poor stars, with magnitudes in the range 10 $ < r \leq 14$. The USS is an ongoing survey; in the near future, it will provide many more bright metal-poor candidate stars for spectroscopic follow-up.

astro-ph.SR↗

arXiv2024

Extending the Chemical Reach of the H3 Survey: Detailed Abundances of the Dwarf-galaxy Stellar Stream Wukong/LMS-1

We present the first detailed chemical-abundance analysis of stars from the dwarf-galaxy stellar stream Wukong/LMS-1 covering a wide metallicity range ($-3.5 < \rm[Fe/H] \lesssim -1.3$). We find abundance patterns that are effectively indistinguishable from the bulk of Indus and Jhelum, a pair of smaller stellar streams proposed to be dynamically associated with Wukong/LMS-1. We confirmed a carbon-enhanced metal-poor star ($\rm[C/Fe] > +0.7$ and $\rm[Fe/H] \sim -2.9$) in Wukong/LMS-1 with strong enhancements in Sr, Y, and Zr, which is peculiar given its solar-level [Ba/Fe]. Wukong/LMS-1 stars have high abundances of $α$ elements up to $\rm[Fe/H] \gtrsim -2$, which is expected for relatively massive dwarfs. Towards the high-metallicity end, Wukong/LMS-1 becomes $α$-poor, revealing that it probably experienced fairly standard chemical evolution. We identified a pair of N- and Na-rich stars in Wukong/LMS-1, reminiscent of multiple populations in globular clusters. This indicates that this dwarf galaxy contained at least one globular cluster that was completely disrupted in addition to two intact ones previously known to be associated with Wukong/LMS-1, which is possibly connected to similar evidence found in Indus. From these $\geq$3 globular clusters, we estimate the total mass of Wukong/LMS-1 to be ${\approx}10^{10} M_\odot$, representing ${\sim}1$% of the present-day Milky Way. Finally, the [Eu/Mg] ratio in Wukong/LMS-1 continuously increases with metallicity, making this the first example of a dwarf galaxy where the production of $r$-process elements is clearly dominated by delayed sources, presumably neutron-star mergers.

astro-ph.GA↗