<?xml version="1.0" encoding="UTF-8"?><xml><records><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>17</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Maharana, Piyush Ranjan</style></author><author><style face="normal" font="default" size="100%">Verma, Ashwini</style></author><author><style face="normal" font="default" size="100%">Joshi, Kavita</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">Retrieval augmented generation for building datasets from scientific literature</style></title><secondary-title><style face="normal" font="default" size="100%">Journal of Physics-Materials</style></secondary-title></titles><keywords><keyword><style  face="normal" font="default" size="100%">dataset building</style></keyword><keyword><style  face="normal" font="default" size="100%">Hydrogen storage</style></keyword><keyword><style  face="normal" font="default" size="100%">LLM</style></keyword><keyword><style  face="normal" font="default" size="100%">materials</style></keyword><keyword><style  face="normal" font="default" size="100%">RAG</style></keyword></keywords><dates><year><style  face="normal" font="default" size="100%">2025</style></year><pub-dates><date><style  face="normal" font="default" size="100%">JUL</style></date></pub-dates></dates><volume><style face="normal" font="default" size="100%">8</style></volume><pages><style face="normal" font="default" size="100%">035006</style></pages><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">&lt;p&gt;
	In this work, we show that employing retrieval augmented generation (RAG) with a large language model (LLM) enables us to extract accurate data from scientific literature and construct datasets. The rapid growth in publications necessitates the automation of extraction of structured data as it is crucial for training machine learning(ML) models. The pipeline developed is simple and can be adjusted accordingly with natural language as input. Quantization enables us to run LLMs on consumer hardware and remove the reliance on closed-source models. Both Llama3-8B and Gemma2-9B with RAG give structured output consistently and with high accuracy as compared to direct prompting. Using the newly developed protocol, we created a data set of metal hydrides for solid-state hydrogen storage from paper abstracts. The accuracy of the generated dataset was &amp;gt;88% in the cases tested. Further, we demonstrate that the generated dataset is ready-to-use for ML models by testing it with HYST to predict the H(2)wt\textbackslash% at a given temperature. Thus, we demonstrate a pipeline to create datasets from scientific literature at minimal computational cost and high accuracy.&lt;/p&gt;
</style></abstract><issue><style face="normal" font="default" size="100%">3</style></issue><work-type><style face="normal" font="default" size="100%">Article</style></work-type><custom3><style face="normal" font="default" size="100%">&lt;p&gt;
	Foreign&lt;/p&gt;
</style></custom3><custom4><style face="normal" font="default" size="100%">&lt;p&gt;
	4.3&lt;/p&gt;
</style></custom4></record><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>17</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Verma, Ashwini</style></author><author><style face="normal" font="default" size="100%">Joshi, Kavita</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">Pathfinder: adaptive learning for hydrogen storage material optimizationa</style></title><secondary-title><style face="normal" font="default" size="100%">International Journal of Hydrogen Energy</style></secondary-title></titles><keywords><keyword><style  face="normal" font="default" size="100%">Adaptive learning</style></keyword><keyword><style  face="normal" font="default" size="100%">Hydrogen storage</style></keyword><keyword><style  face="normal" font="default" size="100%">Metal hydrides</style></keyword><keyword><style  face="normal" font="default" size="100%">PCT isotherms</style></keyword></keywords><dates><year><style  face="normal" font="default" size="100%">2026</style></year><pub-dates><date><style  face="normal" font="default" size="100%">MAY </style></date></pub-dates></dates><volume><style face="normal" font="default" size="100%">236</style></volume><pages><style face="normal" font="default" size="100%">155124</style></pages><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">&lt;p&gt;
	Progress in solid-state hydrogen storage is constrained by time-consuming experiments and scarce high-quality data, limiting effective use of machine learning. To address this, we present an adaptive learning (AL) framework that integrates uncertainty quantification within a closed-loop workflow for targeted material optimization. Unlike static models, it adaptively selects compositions to maximize information gain and improve predictive performance. As a proof of concept, we predict pressure-composition-temperature (PCT) isotherms of Mg-Ni-La systems using literature data for Mg-Ni. The framework identifies informative compositions across Mg fractions (92%-4%) and temperatures (300-633 K), demonstrating effective exploration of the chemical space. For ten unseen compositions evaluated sequentially, accuracy reaches 80% within five cycles, with predictions aligning well with experiments. This establishes a family-specific predictive tool for Mg-Ni-based systems, while the underlying AL framework is broadly applicable to other chemical families with modest initial experimental datasets.&lt;/p&gt;
</style></abstract><work-type><style face="normal" font="default" size="100%">Article</style></work-type><custom3><style face="normal" font="default" size="100%">&lt;p&gt;
	Foreign&lt;/p&gt;
</style></custom3><custom4><style face="normal" font="default" size="100%">&lt;p&gt;
	9.2&lt;/p&gt;
</style></custom4></record><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>17</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Verma, Ashwini</style></author><author><style face="normal" font="default" size="100%">Joshi, Kavita</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">What drives property prediction for solid-state hydrogen storage? data or smart features?</style></title><secondary-title><style face="normal" font="default" size="100%">International Journal of Hydrogen Energy</style></secondary-title></titles><keywords><keyword><style  face="normal" font="default" size="100%">Experimental data</style></keyword><keyword><style  face="normal" font="default" size="100%">Feature engineering</style></keyword><keyword><style  face="normal" font="default" size="100%">Hydrogen storage</style></keyword><keyword><style  face="normal" font="default" size="100%">machine learning</style></keyword><keyword><style  face="normal" font="default" size="100%">Metal hydrides</style></keyword><keyword><style  face="normal" font="default" size="100%">Property prediction</style></keyword></keywords><dates><year><style  face="normal" font="default" size="100%">2026</style></year><pub-dates><date><style  face="normal" font="default" size="100%">APR </style></date></pub-dates></dates><volume><style face="normal" font="default" size="100%">226</style></volume><pages><style face="normal" font="default" size="100%">154499</style></pages><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">&lt;p&gt;
	Metal hydrides play a pivotal role in a wide range of applications, including hydrogen storage, compression, heat management, and catalysis, making them a central focus of interdisciplinary research spanning chemistry, materials science, and engineering. The performance of the metal hydride-based systems is strongly governed by the thermodynamics of metal-hydrogen interactions. Among key thermodynamic properties, the equilibrium plateau pressure (P-eq) is particularly critical, as it defines operating conditions for hydrogen absorption and desorption. Traditionally, determining P-eq requires extensive experimental measurements, which limits the pace of materials discovery. On the other hand, predicting it through ML-based models is constrained by the availability of limited data. In this work, we demonstrate that smart features can be a way to overcome this limitation. EquiP, an ML model trained to predict ln(P-eq) as a function of temperature, generates Van't Hoff plots (P-eq vs. 1/T), enabling rapid determination of enthalpy and entropy of hydride formation. We demonstrate that incorporating structural descriptors derived from X-ray diffraction (XRD) data improves the performance of the model, particularly with sparse training datasets. A model trained using only compositional descriptors yields a validation mean absolute error (MAE) of 0.21 bar, whereas incorporating XRD features reduces the MAE substantially to 0.07 bar. This work demonstrates that with limited data, intelligent feature design grounded in domain knowledge is the key to improving predictions of complex material properties.&lt;/p&gt;
</style></abstract><work-type><style face="normal" font="default" size="100%">Article</style></work-type><custom3><style face="normal" font="default" size="100%">&lt;p&gt;
	Foreign&lt;/p&gt;
</style></custom3><custom4><style face="normal" font="default" size="100%">&lt;p&gt;
	8.3&lt;/p&gt;
</style></custom4></record></records></xml>