{"title":"Can model attribution bridge AI's accountability gap in safety-critical domains?","authors":"Marc Juarez","doi":"10.1098/rsta.2025.0117","DOIUrl":"https://doi.org/10.1098/rsta.2025.0117","url":null,"abstract":"<p><p>The current trend of deploying machine learning models as remote services obscures which specific models users are interacting with, exacerbating an accountability gap that becomes ever more pressing as these technologies make their way into safety-critical domains. In this paper, we examine how model attribution, a recently proposed accountability property, can address this gap. We review recent technical approaches to provide model attribution and identify limitations, such as computational costs, latency overheads and insufficient guarantees, that render them inadequate for many safety-critical applications. We conclude that although model attribution is an important step towards accountability, current technical solutions fall short of the requirements of safety-critical domains. This article is part of the theme issue 'Safe, secure and robust AI for safety-critical systems'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2324","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148472074","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Trustworthiness in AI: on SciCompBot-the scientific computing chatbot.","authors":"Anders Christian Hansen, Fabian Circelli","doi":"10.1098/rsta.2025.0116","DOIUrl":"https://doi.org/10.1098/rsta.2025.0116","url":null,"abstract":"<p><p>This paper explores the prospect of a scientific computing chatbot (SciCompBot): an artificial intelligence (AI) system designed to address advanced problems in optimization, spectral computations (eigenvalues, eigenvectors, spectra), differential equations, inverse problems and related areas. With powerful chatbots like ChatGPT-5.2, it is no longer science fiction for AI to accept these problems; ChatGPT-5.2 does so willingly. Yet, current chatbots suffer from a critical flaw: they often produce incorrect answers with unwarranted confidence, making them fundamentally> untrustworthy for scientific computing. We investigate the mathematical challenges involved in developing a trustworthy SciCompBot and demonstrate that tools from the foundations of computational mathematics, particularly the solvability complexity index (SCI) hierarchy, are necessary in this endeavour. Our results present new strategies for overcoming central obstacles for building a trustworthy SciCompBot: the issues of non-computability and the phenomenon of generalized hardness of approximation (GHA). We identify a key feature that a trustworthy SciCompBot must have: it must be 'chatty'. That is, it must initiate a dialogue with the user. This work fits in the broader programme of ensuring trustworthiness and safety in AI systems, which is now a major topic of interest owing to explosive growth in AI development in recent years. This article is part of the theme issue 'Safe, secure and robust AI for safety-critical systems'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2324","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148472226","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Vulnerability analysis of transformer-based optical character recognition to adversarial attacks.","authors":"Lucas Beerens, Desmond J Higham","doi":"10.1098/rsta.2025.0127","DOIUrl":"https://doi.org/10.1098/rsta.2025.0127","url":null,"abstract":"<p><p>We present a novel framework to assess the resilience of state-of-the-art transformer-based optical character recognition (TrOCR) models. In this way, we develop new untargeted and targeted attack algorithms. On a benchmark handwriting dataset, we show that adversarial perturbations to the input reveal dramatic vulnerabilities without being noticeable to the eye. This article is part of the theme issue 'Safe, secure and robust AI for safety-critical systems'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2324","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148472152","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Neuro-symbolic weak supervision: theory and semantics.","authors":"Nijesh Upreti, Vaishak Belle","doi":"10.1098/rsta.2025.0115","DOIUrl":"https://doi.org/10.1098/rsta.2025.0115","url":null,"abstract":"<p><p>Weak supervision enables machine learning models to learn from limited or noisy labels, but it introduces challenges in reliability and semantic clarity, particularly in multi-instance partial label learning (MI-PLL), where models must resolve both ambiguous supervision signals and uncertain instance-label mappings. This paper proposes a semantics for a neuro-symbolic framework that integrates inductive logic programming (ILP) to structure MI-PLL through relational constraints. In this formulation, ILP defines a hypothesis space over label transitions, formalizes the semantics of per-instance classifiers and provides a relational scaffold for reasoning about weak supervision. Two inductive tasks are studied in this framework: inferring the transition predicate (TP) from the observed and classifier predicates (CPs), and inferring instance-level classifier assignments from the observed and TPs. This formal semantics facilitates constraint specification, consistency checking and the diagnosis of semantic failure modes that bag-level accuracy alone may conceal. No claim is made, however, that it renders the underlying neural classifier more transparent. This article is part of the theme issue 'Safe, secure and robust AI for safety-critical systems'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2324","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148472115","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Signs of sulfur fractionation under high magnetic field strength.","authors":"Dominik Orlovskij, Andy To, David Long","doi":"10.1098/rsta.2025.0180","DOIUrl":"https://doi.org/10.1098/rsta.2025.0180","url":null,"abstract":"<p><p>Sulfur, with a first ionization potential (FIP) of 10.36 eV, lies at the boundary between low- and high-FIP elements, making it particularly sensitive to fractionation processes in the solar atmosphere. Sulfur exhibits variable behaviour across solar environments, with coronal remote sensing studies often observing it as a high-FIP element, while in situ measurements sometimes detect low-FIP-like enhancement. Sulfur also exhibits variable behaviour during flares and magnetic restructuring. To understand sulfur's variations, we quantify how sulfur's FIP bias depends on potential field source surface (PFSS)-derived loop properties. We analyse nine Hinode/EUV Imaging Spectrometer (EIS) raster observations using four diagnostic line pairs (Si x 258.37 Å /S x 264.23 Å, S xi 188.68 Å /Ar xi 188.81 Å, Ca xiv 193.87 Å /Ar xiv 194.40 Å and Fe xvi 262.98 Å /S xiii 256.69 Å), with FIP biases derived using differential emission measures (DEM) calculated via regularized inversion. Our results show that the abundance of low-FIP elements, including sulfur, decreases above approximately 150 G relative to the high-FIP element Ar, while showing no dependence on loop length. This provides evidence that FIP fractionation is modulated by the mean magnetic field strength of the coronal loops. This article is part of the Theo Murphy meeting issue 'Solar atmospheric abundances in space and time'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2323","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148376576","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Andy S H To, J Martin Laming, Jeffrey Reep, Adam J Finley
{"title":"Chromospheric dynamics and turbulence regulate the solar FIP effect.","authors":"Andy S H To, J Martin Laming, Jeffrey Reep, Adam J Finley","doi":"10.1098/rsta.2025.0374","DOIUrl":"https://doi.org/10.1098/rsta.2025.0374","url":null,"abstract":"<p><p>Elemental abundance variations in the solar corona, characterized by first ionization potential (FIP) bias, provide crucial diagnostics of chromospheric processes. The ponderomotive force model, which attributes fractionation to Alfvén wave propagation, has successfully reproduced observed abundance patterns in various solar features. However, existing theoretical implementations rely on a static quiet-Sun chromosphere, leaving the influence of chromospheric dynamics largely unexplored. We address this limitation by combining hydrodynamic simulations from HYDRAD with ponderomotive force calculations through FIPpy, a new open-source code. Comparing predictions between an initial VAL-C chromosphere and a heated chromosphere following impulsive nanoflare-like events, we show that the ponderomotive force model remains consistent under dynamic chromospheric conditions, while stronger changes in fractionation behaviour arise from variations in acoustic flux and turbulence. Most significantly, when the acoustic wave flux drops below approximately 5×106 erg cm-2 s-1, mass-dependent thermal velocities dominate the fractionation process, producing counterintuitive patterns where Fe exceeds Ca in FIP bias, while high-FIP Ar shows fractionation. We demonstrate that any source of chromospheric turbulence will act to suppress fractionation. For flares, our results predict that the increased turbulence will suppress FIP bias, potentially explaining the observed abundance variations during flares. These findings suggest that coronal abundances and composition encode a sensitive balance between dominant mechanisms, determined by the ratio of ponderomotive acceleration to turbulent velocity. This article is part of the Theo Murphy meeting issue 'Solar atmospheric abundances in space and time'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2323","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148376710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Juan Martinez-Sykora, Paola Testa, Deborah Baker, Bart De Pontieu
{"title":"Multi-fluid multi-species models for inverse first ionization potential effect.","authors":"Juan Martinez-Sykora, Paola Testa, Deborah Baker, Bart De Pontieu","doi":"10.1098/rsta.2025.0178","DOIUrl":"https://doi.org/10.1098/rsta.2025.0178","url":null,"abstract":"<p><p>The inverse first ionization potential (FIP) effect is rarely observed in the solar atmosphere, and this anomaly poses a challenging problem in understanding physical processes driving this chemical fractionation. In this work, we investigate various scenarios where the inverse FIP effect could occur using simplified one-dimensional multi-fluid MHD models. The model treats the full MHD equations with multi-fluid and multi-species effects, rather than using wave analysis to derive the ponderomotive force and semi-empirical one-dimensional models. In the parametric study considered here, for upward Alfvén waves, one can achieve a negative (opposite) ponderomotive force when the magnetic field strength and the magnetic flux tubes' expansion with height counteract the dissipation and damping effects from multi-fluid interactions. This article is part of the Theo Murphy meeting issue 'Solar atmospheric abundances in space and time'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2323","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148376600","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Spectropolarimetric analysis of waves linked to first ionization potential.","authors":"Mariarita Murabito, Marco Stangalini","doi":"10.1098/rsta.2025.0184","DOIUrl":"https://doi.org/10.1098/rsta.2025.0184","url":null,"abstract":"<p><p>High-resolution spectropolarimetry has opened new avenues for understanding how chromospheric waves shape coronal plasma composition. All modelling efforts so far highlight wave activity, and in particular the ponderomotive force associated with Alfvénic perturbations, as a key ingredient. Over recent years, studies based on spectropolarimetric measurements in the solar chromosphere have identified magnetic perturbations associated with waves linked to regions in the corona with enhanced first ionization potential (FIP) bias. Building on this established framework, the present work explores the diagnostic potential of Stokes-V amplitude asymmetries as an alternative diagnostic tool for investigating wave behaviour relevant to compositional fractionation processes. This article is part of the Theo Murphy meeting issue 'Solar atmospheric abundances in space and time'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2323","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148376581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Andrea Mesoraca, Marco Stangalini, Fabio Reale, Deborah Baker, David Long, Teodora Mihailescu, Mariarita Murabito, Michele Berretti
{"title":"Exploring the link between wave activity in the photospheric velocity driver and the FIP bias in the solar corona.","authors":"Andrea Mesoraca, Marco Stangalini, Fabio Reale, Deborah Baker, David Long, Teodora Mihailescu, Mariarita Murabito, Michele Berretti","doi":"10.1098/rsta.2025.0373","DOIUrl":"https://doi.org/10.1098/rsta.2025.0373","url":null,"abstract":"<p><p>We investigate wave activity in the solar photosphere within an active region, aiming to assess its possible connection to a well-known phenomenon in the solar corona: the first ionization potential (FIP) effect. This effect produces chemical inhomogeneities in the solar atmosphere, where low-FIP elements become preferentially enriched in upper layers, potentially through interactions with magnetohydrodynamic waves. Although earlier studies examined individual active regions over short intervals, here we extend this approach by tracking the evolution of AR 12665 throughout its full-disc transit. We analyse the behaviour of high-frequency waves in the photospheric velocity field over 6 consecutive days, during which 16 FIP-bias maps were obtained with Hinode/EIS. Our results reveal a recurrent enhancement of oscillatory power in the 4-5 mHz band within the sunspot umbra. The temporal evolution of this enhanced power shows a qualitative correspondence with changes in the coronal FIP-bias distribution. These findings suggest a possible link between wave activity in the lower solar atmosphere and elemental fractionation in the corona. By following the active region over an extended period, we show that both the photospheric wave signatures and the coronal composition display temporal evolution, providing observational support for a connection between photospheric dynamics and coronal abundances. This article is part of the Theo Murphy meeting issue 'Solar atmospheric abundances in space and time'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2323","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148376095","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Comparing first ionization potential bias diagnostics in the solar atmosphere.","authors":"Kristena D Spruksta, David M Long, Andy S H To","doi":"10.1098/rsta.2025.0172","DOIUrl":"https://doi.org/10.1098/rsta.2025.0172","url":null,"abstract":"<p><p>Plasma composition in the solar atmosphere differs between the photosphere and corona, producing an observable difference in elemental abundance known as the first ionization potential (FIP) effect. The FIP effect is characterized by the ratio of low to high FIP elements, giving a number known as the FIP bias. FIP bias values vary between different regions of the solar atmosphere, with typical observed values of approximately 1 for coronal holes, approximately 1.5-2 for the quiet Sun and approximately 3 for active regions. The Extreme ultraviolet Imaging Spectrometer (EIS) onboard the Hinode spacecraft has enabled the widespread use of the Si X/S X line pair as a FIP bias diagnostic, but EIS observes other line pairs that can be used to estimate FIP bias. We consider three FIP bias diagnostics observed by Hinode/EIS (Si X/S X, Ca XIV/Ar XIVand Fe XVI/S XIII), comparing the FIP bias between the quiet Sun and an active region. We also assume a range of signal-to-noise ratio (SNR) cut-off values for each pixel, finding that while the SNR cut-off affects the number of useable pixels, higher (lower) SNR cut-offs remove (retain) a tail of high FIP bias values within the measured distribution. However, the median value of the FIP bias distribution remains largely unchanged. These results show the importance of a more nuanced view of FIP bias when using this vitally important diagnostic rather than a simplistic one-size-fits-all approach. This article is part of the Theo Murphy meeting issue 'Solar atmospheric abundances in space and time'.</p>","PeriodicalId":19879,"journal":{"name":"Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences","volume":"384 2323","pages":""},"PeriodicalIF":3.7,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148376695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}