Indian Institute of Technology (Indian School of Mines), Dhanbad (abbreviated IIT (ISM), Dhanbad) is a public technical university located in Dhanbad, India. It has main campus of 218 acres in Sardar Patel Nagar area of Dhanbad and recently Jharkhand Cabinet approved 226.98 acres of land in Nirsa for its second campus, IIT (ISM) Dhanbad administration has plans to open more than 29 academic centers in this upcoming campus in Nirsa, Dhanbad. It is an Institute of National Importance. IIT (ISM) has 18 academic departments covering Engineering, Applied Sciences, Humanities and Social Sciences and Management programs. It was formerly known as Indian School of Mines, Dhanbad before its conversion into an Indian Institute of Technology (IIT).IIT (ISM) Dhanbad is located in the mineral-rich region of India, in the city of Dhanbad. It is the third oldest institute (after IIT Roorkee, and IIT (BHU) Varanasi) which got converted into an IIT. It was established by British Indian Government on the lines of the Royal School of Mines - London, and was formally inaugurated on 9 December 1926 by Lord Irwin, the then Viceroy of India. It started as an institution to impart education in mining and mineral sciences, and today, has grown into a technical institution with various academic departments. IIT (ISM) Dhanbad admits its undergraduate students through Joint Entrance Examination (Advanced), previously IIT-JEE.On 25 May 2016, the Union Cabinet headed by Prime Minister Modi gave its approval to amend the Institutes of Technology Act, 1961 for conversion of ISM Dhanbad into an Indian Institute of Technology. The amendment was approved by parliament and upon presidential assent, was notified in the Gazette of India on 10 August 2016.
The increasing reliance on artificial intelligence (AI) for societal needs presents a growing environmental concern, particularly regarding user perceptions and accountability. This study examines how Young Indian Adults (YIAs; aged 18–24) reconcile their environmental values with their reliance on AI and assesses their perceived locus of accountability for the environmental repercussions of large language models (LLMs) within hybrid human–AI systems. A cross-sectional, open-ended survey (N = 349) was administered to students at an Indian technical university. The survey featured a three-phase inquiry, including an informational prompt detailing the environmental costs of LLMs. The Machine-Assisted Topic Analysis (MATA) framework was used to analyze the data by combining dictionary-based lexicon-matching with Latent Dirichlet Allocation (LDA) topic modeling. Baseline discourse showed an epistemic gap over material costs, which is consistent with the theory of sociotechnical abstraction, where the material reality of data is obscured. However, the informational stimulus was followed by a discursive split, changing the focus toward concerns pertaining to energy consumption and carbon emissions of LLMs. This study identified a discursive inversion of accountability. YIAs attribute the environmental costs of LLMs exclusively to the tech industry, despite acknowledging personal and civic responsibility for climate change in general. This challenges Western-centric AI ethics by demonstrating that environmental awareness among tech-educated youth in India manifests as calls for systemic efficiency rather than behavioral abstinence. To minimize computational opacity, the study argues for shifting policy from digital to ecological sovereignty and shifting education toward material AI literacy.
Vision-Language Models (VLMs) increasingly power high-stakes applications, from medical imaging to autonomous systems, yet they routinely hallucinate, confidently describing content not present in the input. We investigate the root causes of these failure modes with a mechanistic analysis focusing on the decoder-based VLMs. We trace these failure modes to a geometric over-alignment: to bridge the modality gap required by attention mechanisms, decoder-based VLMs over-align visual embeddings with the text manifold, injecting a statistical linguistic bias that systematically overshadows fine-grained visual evidence. While prior work either aggressively closes this gap or suppresses hallucinations through expensive black-box decoding strategies, none addresses the underlying geometric cause. We provide the first quantitative characterization of this over-alignment, demonstrating that linguistic bias concentrates in the top principal components of a universal, dataset-agnostic text subspace. Building on this insight, we propose two complementary remedies: a training-free inference strategy and a bias-aware fine-tuning paradigm, both of which explicitly project out this subspace from visual representations. Our methods significantly reduce hallucinations across POPE, CHAIR, and AMBER benchmarks, and improve CLAIR scores on long-form captioning tasks, with the training-free variant adding no computational overhead over the base model.
The advance rate (AR) of a tunnel boring machine (TBM) governs construction scheduling, cost control, and overall project efficiency; thus, its accurate prediction is essential for effective resource allocation and mitigation of delays arising from geological and operational variability. This study develops an optimal soft-computing framework by comparatively evaluating support vector regression (SVR), feedforward neural networks (FFNN), gene expression programming (GEP), gated recurrent units (GRU), long short-term memory (LSTM), and bidirectional LSTM (BiLSTM) models. A dataset comprising 1,197 TBM operational records was utilized, incorporating cutterhead rotation speed (CRS), mean thrust (F/A), mean cutterhead torque (T/D³), upper earth pressure (UEP), lower earth pressure (LEP), and torque penetration index (TPI). Multicollinearity among predictors was quantified using the variance inflation factor (VIF), while feature sensitivity was assessed via the cosine amplitude method. Model performance was evaluated using eight statistical indices, three reliability measures, regression error characteristic (REC) curves, generalizability assessment, and the Wilcoxon signed-rank test. Comparative analysis demonstrated the superior predictive capability of the BiLSTM model, achieving accuracy exceeding 98.60
A compact two-dimensional photonic crystal (2D-PhC) slab is proposed, consisting of silicon rods arranged in a hexagonal lattice within an air background, designed for biosensing applications. The photonic band structure is analyzed by the plane-wave expansion technique, while the sensing performance is evaluated through the finite-difference time-domain method. To assess its capabilities, the transmission spectra and the resulting resonance wavelength shifts are thoroughly analyzed. Saliva and glucose detection are demonstrated by monitoring refractive index variations at different concentrations with a 1550 nm laser source. In addition, the effect of shifting the silicon rods, highlighted with a red circle and horizontal bar symbol, is carefully examined to optimize the sensor response. The structure is examined in two-dimensional (2D) and threedimensional (3D) configurations, highlighting sensor performance. For saliva detection, the sensor achieves a high quality factor of 1.4333 x 105, a maximum sensitivity of 883 nm/RIU, a figure of merit of 8.3926 x 104 RIU-1, and a detection limit of 1.192 x 10-6 RIU along with an average response time of approximately 219 fs. For glucose detection, it achieves a quality factor of 1.5915 x 105, a sensitivity of 766.67 nm/RIU, a figure of merit of 6.755 x 104 RIU-1, and a detection limit of 1.48 x 10-6 RIU. Furthermore, the device demonstrates reliable operation across a wide temperature range (0-50 degrees C), and the impact of fabrication tolerances on its performance is comprehensively examined. Finally, its compact footprint of 130.235 mu m2 highlights its suitability for integration into optical microchips and diverse biosensing applications.
Phosphorites constitute a potential non-traditional source of rare earth elements (REE), although their resource viability is still being evaluated globally. This study presents the first documentation of REE-enriched sedimenthosted phosphorites from India, and evaluates the processes driving REE mobilization, enrichment and accumulation. The Ediacaran phosphatic beds from the Owk Shale Formation, Kurnool Group of the Cuddapah basin, interbedded with shales and tuffs, were analysed using optical and SEM imaging, whole rock geochemical, in-situ fluorapatite (EPMA, LA-ICP-MS) and Raman spectroscopic analyses. Two apatite phases were identified: primary authigenic (Ap-1), fluorine-poor variety and secondary/reprecipitated REE-enriched fluorapatite (Ap-2). Wholerock REE + Y concentrations reach up to 1075 ppm, with characteristic middle REE (MREE) enriched patterns, also recorded in fluorapatite, interpreted as a result of diagenetic uptake from sediment porewaters. A strong correlation between REE + Y and redox tracer, Mn indicates initial MREE scavenging by Mn-(oxy)hydroxides and organic matter, followed by MREE release during Mn-reductive dissolution. While diagenesis enhanced REE enrichment, post-diagenetic hydrothermal reworking associated with igneous intrusions diluted REE concentrations through the formation of abundant Fe-Mg chlorite (chamosite) and REE and P redistribution within the interbedded shales. Sequence stratigraphic analysis suggests that these pristine phosphorites were deposited during marine regression, without subsequent reworking, preventing the formation of high grade phosphorite deposit, emphasizing the role of sea-level changes in controlling phosphogenesis, formation of a phosphorite giant and thus, REE enrichment. A temporal analysis of global phosphorites further reveals that progressive seawater influence, modulated by biogeochemical recycling and episodic weathering, together controlled the REE enrichment patterns in phosphorites through Earth's history.
