01743nas a2200181 4500000000100000000000100001000000100002008004100003260000600044100001900050700002100069700002200090700001800112245009100130490000700221520131900228022001401547 2015 d c91 aPawel Foremski1 aMichał Gorawski1 aKrzysztof Grochla1 aKonrad Połys00aEnergy-Efficient Crowdsensing of Human Mobility and Signal Levels in Cellular Networks0 v153 aThe paper presents a practical application of the crowdsensing idea to measure human mobility and signal coverage in cellular networks. Currently, virtually everyone is carrying a mobile phone, which may be used as a sensor to gather research data by measuring, e.g., human mobility and radio signal levels. However, many users are unwilling to participate in crowdsensing experiments. This work begins with the analysis of the barriers for engaging people in crowdsensing. A survey showed that people who agree to participate in crowdsensing expect a minimum impact on their battery lifetime and phone usage habits. To address these requirements, this paper proposes an application for measuring the location and signal strength data based on energy-efficient GPS tracking, which allows one to perform the measurements of human mobility and radio signal levels with minimum energy utilization and without any engagement of the user. The method described combines measurements from the accelerometer with effective management of the GPS to monitor the user mobility with the decrease in battery lifetime by approximately 20%. To show the applicability of the proposed platform, the sample results of signal level distribution and coverage maps gathered for an LTE network and representing human mobility are shown. a1424-822001010nas a2200133 4500000000100000000000100001000000100002008004100003100001800044245007700062300001200139490000700151520071800158 2014 d1 aKonrad Połys00aWpływ wielkości histerezy przełączeń na przepustowość w sieci LTE a53–620 v353 aW sieciach mobilnych użytkownicy są przełączani pomiędzy stacjami bazowymi na podstawie siły sygnału lub innych metryk jakości transmisji. W celu optymalizacji przełączeń pomiędzy stacjami bazowymi często wykorzystywana jest histereza. Pozwala ona ograniczyć liczę niepotrzebnych, bardzo krótkich przełączeń. W artykule zbadano wpływ wielkości histerezy na liczbę przełączeń oraz na osiąganą przepustowość połączenia użytkownika sieci LTE wyznaczaną na podstawie poziomu sygnału. Wyniki przeprowadzonych symulacji wskazują, że odpowiedni margines przełączenia znacznie zmniejsza liczbę niepotrzebnych przełączeń, a jednocześnie praktycznie nie zmniejsza przepustowości.01116nas a2200145 4500000000100000000000100001008004100002260005000043100001800093700002200111700001400133245007500147300001400222520073400236 2015 d bSpringer International Publishing Switzerland1 aKonrad Połys1 aKrzysztof Grochla1 aPiotr Gaj00aMobility robustness in LTE based on automatic neighbor relation table. a232–2413 aWe evaluate the effect of automatic updates to the LTE Neighbor Relation Table (NRT) to optimize the efficiency of handovers in the network. We propose an algorithm which monitors the handovers within the LTE network and detects events when the procedure was started to early (handover to early) or too late (handover too late). Basing on the observed statistics in Operation and Maintenance server (OAM) the proposed algorithm updates the NRT to maximize the network throughput and minimize the probability of loss of connectivity. Performed simulations show that correctly forbidden handovers can reduce number of unnecessary signaling, reduce risk of fail during handover and reduce the time required to perform the handovers.00271nas a2200121 4500000000100000000000100001008004100002260001300043100002200056700001800078245003900096300001400135 2016 d bSpringer1 aKrzysztof Grochla1 aKonrad Połys00aUsing Kalman Filters on GPS Tracks a663–67200306nas a2200109 4500000000100000000000100001008004100002260001000043100002200053700001800075245010300093 2015 d bKSTiT1 aKrzysztof Grochla1 aKonrad Połys00a{Dobór parametrów mechanizmu zwielokrotnienia wykorzystania częstotliwości SFR w sieciach LTE}00557nas a2200145 4500000000100000000000100001008004100002260013200043100002200175700002100197700001800218245010100236856005200337020002200389 2015 d bICST (Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering)aICST, Brussels, Belgium, Belgium1 aKrzysztof Grochla1 aMichał Gorawski1 aKonrad Połys00aComparison of Measurements and Simulation Model of Indoor Signal Level Variation in LTE Networks uhttp://dx.doi.org/10.4108/eai.24-8-2015.2261125 a978-1-63190-079-200312nas a2200121 4500000000100000000000100001008004100002260001300043100002200056700001800078245008000096300001400176 2016 d bSpringer1 aKrzysztof Grochla1 aKonrad Połys00aSubcarrier Allocation for LTE Soft Frequency Reuse Based on Graph Colouring a447–45400268nas a2200109 4500000000100000000000100001008004100002100002100043700002200064700001800086245005400104 2015 d1 aMichał Gorawski1 aKrzysztof Grochla1 aKonrad Połys00a{Variability of LTE Signal in Indoor Environment}00492nas a2200193 4500000000100000000000100001008004100002260001300043100002900056700002100085700002200106700002800128700001800156700001900174700001500193700001600208245006100224300001300285 2014 d bSpringer1 aFrancisco Barcelo-Arroyo1 aMichał Gorawski1 aKrzysztof Grochla1 aIsrael Martín-Escalona1 aKonrad Połys1 aAndrea Ribeiro1 aRute Sofia1 aEnrica Zola00aNew Trends in Mobility Modelling and Handover Prediction a88–11400376nas a2200157 4500000000100000000000100001008004100002260001300043100002200056700001600078700001100094700002200105700001800127245005900145300001400204 2014 d bSpringer1 aGiovanni Giambene1 aTara Yahiya1 aVan Le1 aKrzysztof Grochla1 aKonrad Połys00a{Resource Management and Cell Planning in LTE Systems} a177–19700270nas a2200097 4500000000100000000000100001008004100002100002200043700001800065245008900083 2014 d1 aKrzysztof Grochla1 aKonrad Połys00aSimplied model of wireless network throughput changes resulting from client mobility00277nas a2200097 4500000000100000000000100001008004100002100002200043700001800065245009600083 2013 d1 aKrzysztof Grochla1 aKonrad Połys00aInfluence of the Handoff Threshold Hysteresis on Heterogeneous Wireless Network Performance00311nas a2200109 4500000000100000000000100001008004100002260001000043100002200053700001800075245010800093 2016 d bKSTiT1 aKrzysztof Grochla1 aKonrad Połys00aAnaliza wpływu mocy nadawania stacji bazowych na mechanizmy równoważenia obciążenia w sieciach LTE00313nas a2200109 4500000000100000000000100001008004100002260002000043100001800063700002200081245010000103 2017 d c09/2017aWarsaw1 aKonrad Połys1 aKrzysztof Grochla00aModel zmienności przepływności bitowej transmisji w sieciach LTE oparty o łańcuchy Markowa01184nas a2200145 4500000000100000000000100001008004100002260003800043100001800081700002200099700002100121245008100142520079300223020002201016 2018 d bSpringer International Publishing1 aKonrad Połys1 aKrzysztof Grochla1 aMichał Gorawski00aLTE Load Balancing with Tx Power Adjustment and the Real Life Mobility Model3 a

The load balancing in the LTE and LTE-A networks allows usage of the radio resources more effectively to improve client satisfaction. The paper provides analysis of the influence of complex user mobility patterns on the load balancing algorithm. The simulation models are used to evaluate the number of unsatisfied clients depending on the mobility model used. Two mobility models are used: RLMM and Random Waypoint. The Tx Power Adjustment method is evaluated, which is based on reconfiguration of a LTE eNodeB TX Power and was proposed by authors in their previous papers. Method proved to be effective and fairly easy in implementation, and the evaluation through series of simulation shown up to 50{%} rise in clients satisfaction with upholding the total network efficiency.

 a978-3-319-92459-500431nas a2200157 4500000000100000000000100001008004100002260002200043100001800065700002200083700002100105700002200126245006300148300001200211490005000223 2019 d c06/2019bSpringer1 aKonrad Połys1 aKrzysztof Grochla1 aMichał Gorawski1 aArtur Frankiewicz00aAccuracy of UWB Location Tracking Devices When on the Move a233-2410 vInternational Conference on Computer Networks00307nas a2200121 4500000000100000000000100001008004100002100002200043700001900065700001800084700001700102245006600119 2019 d1 aKrzysztof Grochla1 aRafał Marjasz1 aKonrad Połys1 aAnna Strzoda00aKolizje pakietów w sieciach LoRa w zastosowaniach smart city01346nas a2200133 4500000000100000000000100001008004100002260002000043100002200063700001800085245008400103856005800187520096700245 2020 d c07/2020aCyprus1 aKrzysztof Grochla1 aKonrad Połys00aHeuristic Algorithm for Gateway Location Selection in Large Scale LoRa Networks uhttps://ieeexplore.ieee.org/abstract/document/91484353 a

The LoRa networks allow building low-cost longrange networks. The selection of LoRa access point locations in large scale networks requires taking into consideration the spatial distribution of clients, radio signal propagation and the limitation of the number of devices served by a single point of access. We propose a heuristic algorithm for the selection of access point locations within for a given set of candidate locations, based on gradient optimization. The proposed method allows selecting a sub-optimal set of locations which provide full coverage and take into consideration the capacity dimensioning based on spreading factor and expected channel utilization. The algorithm is evaluated in random topologies and in reallife scenario based on city-wide smart meters locations. The analysis shows that it outperforms a manual selection of access points locations, allowing to decrease the number of access points providing the full coverage.

01441nas a2200205 4500000000100000000000100001008004100002260002200043100002200065700001400087700002200101700002700123700002200150700001800172700002500190700001700215245008000232856003700312520088600349 2020 d c01/2021bSpringer1 aArtur Frankiewicz1 aAdam Glos1 aKrzysztof Grochla1 aZbigniew Łaskarzewski1 aJaroslaw Miszczak1 aKonrad Połys1 aPrzemysław Sadowski1 aAnna Strzoda00aLP WAN Gateway Location Selection Using Modified K-Dominating Set Algorithm uhttps://arxiv.org/abs/2010.045533 a

The LP WAN networks use gateways or base stations to communicate with devices distributed on large distances, up to tens of kilometres. The selection of optimal gateway locations in wireless networks should allow providing the complete coverage for a given set of nodes, taking into account the limitations, such as the number of nodes served per access point or required redundancy. In this paper, we describe the problem of selecting the base stations in a network using the concept of k-dominating set. In our model, we include information about the required redundancy and spectral efficiency. We consider the additional requirements on the resulting connections and provide the greedy algorithm for solving the problem. The algorithm is evaluated in randomly generated network topologies and using the coordinates of sample real smart metering networks.

01795nas a2200181 4500000000100000000000100001008004100002260002700043100001900070700001700089700001800106700002200124245006500146856011600211300001200327520125200339022002201591 2021 d c10/2021bEurosisaRome1 aRafał Marjasz1 aAnna Strzoda1 aKonrad Połys1 aKrzysztof Grochla00aMitigation of LoRa interferences via dynamic channel weights uhttps://www.researchgate.net/publication/356041519_Mitigation_of_LORA_Interferences_via_Dynamic_Channel_Weights a150-1543 a

The Low Power Wide Area Networks, such as LoRa (Long Range), allow communication over a large area with limited bandwidth and small data rates, using an unlicensed band. LoRa uses multiple channels, but the number of channels available is relatively narrow - e.g. 8 in most countries in Europe. The transmission is subject to interferences due to many other devices transmitting within the same frequencies. In this paper, we evaluate how strong the interferences' influence is by analysing the packet delivery rate in a large scale LoRa network. The measurements have been made over more than seven months and show noticeable variability in packet delivery rate per channel both in space and in time. We propose to estimate the scale of interferences through continuous observation of the number of delivered packets for every channel per LoRa gateway or per selected area. We also discuss how this estimation may mitigate the interferences by assigning the dynamic weights to the LoRa channel selection procedure. We created an algorithm that periodically updates the weights, and a simulation model to evaluate the algorithm. A significant improvement in the efficiency of package delivery is visible, from 12% to 49% less packet loss.

 a978-9-492-85918-101878nas a2200169 4500000000100000000000100001008004100002260005900043100001900102700001800121700001700139700002200156245004500178856006700223520140000290020001801690 2021 d bAssociation for Computing MachineryaNew York, NY, USA1 aRafał Marjasz1 aKonrad Połys1 aAnna Strzoda1 aKrzysztof Grochla00aImproving Delivery Ratio in LoRa Network uhttps://dl.acm.org/doi/10.1145/3479241.3486698?cid=996599133973 a

The LoRa (Long Range) radio communication provides large coverage areas, low bandwidth, small data rates, and transmitting limited data sizes, with very low energy usage. Multiple transmission channels are used, but the number of channels available is fairly limited - e.g. 8 in most countries in Europe. The amount of interferences may be estimated by the continuous observation of the number of delivered packets and the received signal strength. We propose to aggregate and analyse those metrics per channel and geographical area. When the proportion of received packets per some of the areas decreases, it is subjected to heavier interferences than other parts of the network and should be used less intensively. Thus we propose to allocate the weights for each channel per geographic area and/or device, defining the probability of the channel usage. Furthermore, we created an algorithm that periodically updates the weights and a simulation model using OMNeT++, FloRa, taking into account possible packets collisions. The algorithm has been validated for both random and realistic network topologies. The simulation evaluations show a significant improvement in the package delivery efficiency, by almost 50% in scenarios with very high interference level simulated and by more than 13% in scenarios fed with real-life measurements of packet delivery probability changes in time.

 a978145039079801805nas a2200157 4500000000100000000000100001008004100002260006700043100001700110700002200127700001800149245012800167856005500295520127900350020001801629 2022 d c10/2022bAssociation for Computing MachineryaMontreal, Canada1 aAnna Strzoda1 aKrzysztof Grochla1 aKonrad Połys00aVariability of BLE Advertisement Packets Received Signal Strength and Delivery Probability in the Presence of Interferences uhttps://dl.acm.org/doi/abs/10.1145/3551662.35609333 a

The Bluetooth Low Energy (BLE) advertisement messages allow estimating the location and distance between mobile devices. We analyse how the signal level varies depending on interferences, distance and the orientation of the transmitting and receiving devices. To provide a representative BLE Received Signal Strength Index (RSSI) dataset, we present a series of experimental measurements, including different receiver antenna orientations for several distances. We analyse the signal level variability as a function of distance and angle in an indoor environment. Our measurements show that the mean RSSI value equals -65.5 dBm, -78.4 dBm, -84.1 dBm and -90.8 dBm for 1m, 5m, 10m, and 16m, respectively. The standard deviation changes from 5.7 dBm through 5.9 dBm, 4.5 dBm to 3.5 dBm for 1m, 5m, 10m, and 16m, respectively. We also investigate the influence of the interference, emulating environment in which devices actively transmit BLE signals with varying intensities; on average, at every 0ms (no additional signals), 4ms and 8ms. The experiments show that from 13% to 40% fewer packets are delivered per second in the presence of interferences. Nevertheless, the RSSI characteristics remain very similar regardless of the presence of emulated traffic or not.

 a978145039482601648nas a2200133 4500000000100000000000100001008004100002260003300043100002100076700001800097245014200115856005000257520120700307 2023 d c03/2023bIEEEaHainan, China1 aMichał Gorawski1 aKonrad Połys00aLocalization of mobile nodes in graphs using the True-range multilateration with node sorting (TriSort) approach in an indoor environment uhttps://ieeexplore.ieee.org/document/100746423 a
Mobile device localization is still an actual research problem. In many cases, we have a lack in all the positions but there is enough data to calculate position with the use of the multilateration. Our research proves that the order in which we calculate the positions with multilateration matters. We consider a network with mobile and stationary devices which are represented in the form of a graph with calculated distances as edges. Our contribution to solving the problem of indoor localization is true-range multilateration with a node sorting approach (TriSort). We based our research on an existing problem of smartphone localization in an enclosed indoor space using the BLE connection. The proposed algorithm is evaluated in a simulator. The distances are calculated using the BLE signal level estimation. We use the Genetic Algorithm commonly used for such problems to assess our TriSort approach. We compare the results of both methods according to obtained errors and their processing times, showing that the proposed TriSort method provides a minimal decrease in accuracy compared to global optimization (below 14%), with almost four orders of magnitudes lower complexity.
00446nas a2200145 4500000000100000000000100001008004100002260001200043100002200055700002200077700001800099245013000117856004600247490000700293 2023 d c11/20231 aViacheslav Kovtun1 aKrzysztof Grochla1 aKonrad Połys00aInvestigation of the Information Interaction of the Sensor Network end IoT Device and the Hub at the Transport Protocol Level uhttps://www.mdpi.com/2079-9292/12/22/46620 v1200512nas a2200157 4500000000100000000000100001008004100002260001200043100002200055700002200077700001900099700001800118245016500136856004600301490000700347 2023 d c12/20231 aViacheslav Kovtun1 aKrzysztof Grochla1 aWojciech Kempa1 aKonrad Połys00aReliably Controlling Massive Traffic between a Sensor Network End Internet of Things Device Environment and a Hub Using Transmission Control Protocol Mechanisms uhttps://www.mdpi.com/2079-9292/12/24/49200 v1200459nas a2200145 4500000000100000000000100001008004100002260001100043100002200054700002200076700001800098245011800116856007200234490000700306 2024 d c5/20241 aViacheslav Kovtun1 aKrzysztof Grochla1 aKonrad Połys00aThe concept of network resource control of a 5G cluster focused on the smart city's critical infrastructure needs uhttps://www.sciencedirect.com/science/article/pii/S11100168240028250 v9400500nas a2200157 4500000000100000000000100001008004100002260001100043100002200054700001800076700002200094700001800116245015200134856004900286490000700335 2024 d c3/20241 aViacheslav Kovtun1 aOksana Kovtun1 aKrzysztof Grochla1 aKonrad Połys00aAssessment of Communication Resource Allocation by the Transmission Control Protocol for the Target Virtual Connection under Competitive Conditions uhttps://www.mdpi.com/2079-9292/13/7/1180/pdf0 v1302017nas a2200157 4500000000100000000000100001008004100002260001200043100001900055700002200074700001800096245013600114856005500250490000700305520154700312 2024 d c06/20241 aRafał Marjasz1 aKrzysztof Grochla1 aKonrad Połys00aA Comprehensive Algorithm for Vertical Positioning in Multi-Building Environments as an Advancement in Indoor Floor-Level Detection uhttps://www.nature.com/articles/s41598-024-64824-90 v143 a

The proliferation of smartphones has catalyzed diverse services, mainly focusing on indoor localization to determine users’ and devices’ positions within buildings. Despite decades of exploration, the seamless integration of wireless technologies in tracking devices and users has become pivotal in various sectors, including health, industry, disaster management, building operations, and surveillance. Extensive research in laboratory and industrial settings, particularly in wireless sensor networks and robotics, has informed indoor localization techniques. This paper, referencing surveys and original literature reviews, proposes an innovative indoor location system amalgamating GPS and barometer readings. GPS identifies entry through doors, while barometer readings facilitate accurate floor-level tracking. The integration promises continuous real-time location updates, enhancing security, navigation, and emergency response. Notably, the algorithm is infrastructure-independent, relying on the smartphone’s barometer, and versatile, detecting elevator travel when Wi-Fi AP or LTE signals are available. Results indicate high accuracy, with building entry exceeding 93%, elevator recognition achieving 75% sensitivity and 97% specificity, and floor change detection surpassing 95% sensitivity and nearly 98% specificity (which translates to nearly 97% accuracy). This comprehensive solution, emphasizing the critical role of precise vertical positioning, signifies an advancement in tracking within urban structures.

01311nas a2200145 4500000000100000000000100001008004100002260000900043100001900052700002000071700001800091245008200109856006500191520090900256 2025 d c20251 aMarzena Halama1 aSławomir Nowak1 aKonrad Połys00aSemantic Orientation for Indoor Navigation System using Large Language Models uhttps://link.springer.com/article/10.1038/s41598-025-25281-03 a

Autonomous robots play an important role in modern indoor navigation, but existing systems often struggle with seamless human interaction and semantic understanding of environments. This paper presents an Artificial Intelligence (AI)-driven object recognition system enhanced by Large Language Models (LLMs), such as GPT-4 Vision and Gemini, to bridge this gap. Our approach combines vision-based mapping techniques with natural language processing and interactions to enable intuitive collaboration in solving navigation tasks. By leveraging multimodal input and vector space analysis, our system achieves enhanced object recognition, semantic embedding, and context-aware responses, setting a new standard for autonomous indoor navigation. This approach provides a novel framework for improving spatial understanding and dynamic interaction, making it suitable for complex indoor environments.

01702nas a2200133 4500000000100000000000100001008004100002260004600043100001900089700001800108700002100126245011700147520130400264 2025 d c2025bIEEE BigData 2025aMacau SAR, China1 aMarzena Halama1 aKonrad Połys1 aJoanna Domańska00aSpeech-based analysis of aggression in social content: resistance of Large Language Models to signal degradation3 a

Artificial Intelligence (AI) systems, including robotic assistants and voice interfaces, are increasingly present in public spaces such as airports, train stations, hospitals, and government offices, offering new opportunities to enhance user safety and comfort. Detecting aggression and hate speech in audio data from such locations is critical to protecting users and creating safe spaces. Previous approaches in this area have mainly relied on classification using classical Machine Learning (ML) methods and Deep Learning (DL) techniques, primarily focused on pure text analysis. However, most systems do not take into account real-world challenges such as background noise, incomplete utterances, or transcription errors resulting from automatic speech recognition. In this work, we propose a comprehensive approach that exploits the potential of Large Language Models (LLMs) to classify aggressive content in degraded audio transcriptions. We create a realistic test environment by converting tagged social media posts with offensive content into recordings with simulated noise and compare the performance of the models under different data quality conditions. Our research provides practical guidance on how to build more resilient and accessible voice moderation systems using LLM.

01795nas a2200253 4500000000100000000000100001000000100002008004100003260002500044100002600069700001800095700002500113700001700138700001800155700002000173700002400193700001800217700001600235245007000251856005100321300001200372490000700384520115000391 2025 d bIOS PressaAmsterdam1 aInna Skarga-Bandurova1 aJames Lambert1 aTetiana Biloborodova1 aSaif Nabhani1 aTiana Kaleeva1 aAndrii Davydiuk1 aVolodymyr Artemchuk1 aKonrad Połys1 aIgor Linkov00aDevelopmental Goals and AI in the Context of Crises and Disasters uhttps://ebooks.iospress.nl/volumearticle/76753 a159-2020 v673 a

Artificial Intelligence (AI) holds significant promise for advancing peace, digital resilience, and sustainable development in countries experiencing conflict, crisis, or undergoing political transition. Grounded in the Sustainable Development Goals (SDGs), this chapter presents a framework for prioritizing SDGs across the phases of crisis (emergency response, stabilization, and long-term development) and demonstrates how AI and digital technologies can operationalize sustainable development even in the most adverse conditions. By shifting from a peace-time ideal to a crisis-informed implementation model, we can transform the SDGs from aspirational blueprints into tools for resilience, recovery, and transformation. We argue that effective AI-enabled resilience begins with robust foundational infrastructure and services, upon which increasingly sophisticated digital security and decision-support layers must be built. Beyond technical measures, we emphasize the need for strategic cross-sector alignment, inclusive governance, and international cooperation coupled with a culture of cybersecurity awareness and preparedness.

00964nas a2200349 4500000000100000000000100001008004100002260002500043100001800068700002000086700002000106700002100126700002500147700002400172700002600196700001600222700001600238700001600254700001800270700001800288700001800306700001700324700002100341700002000362700002000382700001900402700001600421245010500437856005100542300001400593490000700607 2025 d bIOS PressaAmsterdam1 aJames Lambert1 aKelsey Stoddard1 aMegan Marcellin1 aThomas Polmateer1 aTetiana Biloborodova1 aVolodymyr Artemchuk1 aInna Skarga-Bandurova1 aJenny Laird1 aMoshe Vardi1 aBoyan Jekov1 aTiana Kaleeva1 aKonrad Połys1 aPetar Tsankov1 aThomas Welsh1 aEmanuele Bellini1 aErnesto Damiani1 aAndrii Davydiuk1 aBenjamin Trump1 aIgor Linkov00aRoles for AI/ML in Resilience of Cyber-Based Systems and Other Infrastructure: Working Group Summary uhttps://ebooks.iospress.nl/volumearticle/76752 a139 - 1580 v6701663nas a2200121 4500000000100000008004100001260007000042100001900112700001800131700002100149245008600170520128500256 2026 d b2nd International Conference PUT STEM Day 2026: Book of Abstracts1 aMarzena Halama1 aKonrad Połys1 aJoanna Domańska00aAI-Assisted Zeolite Design through Retrieval-Augmented Scientific Language Models3 a

Designing materials such as zeolites is a major challenge due to the wide range of possible structures and synthesis conditions. Furthermore, knowledge on this subject is widely scattered across numerous publications and experimental notes. Therefore, large language models (LLMs) represent a promising tool for building knowledge bases, supporting the analysis of scientific literature, although their responses may be incomplete or fallible. In addition, in the case of zeolites, the relationship between material properties and their topology is of crucial importance, which limits the effectiveness of approaches based solely on text. In this study, we investigate a Retrieval-Augmented Generation (RAG) approach extended with a multimodal component that combines textual and structural information.

 

 

 

00343nas a2200121 4500000000100000008004100001260002500042100001900067700001800086700002000104700002100124245007600145 2026 d bAGH University Press1 aMarzena Halama1 aKonrad Połys1 aSławomir Nowak1 aJoanna Domańska00aExternal Semantic Memory as a Cognitive Layer for Large Language Models