Abstract

Automated table structure recognition from input images faces significant visual and linguistic challenges, including layout variances, table densities, and the presence of empty or multi-line cells. This problem is crucial due to its extensive applications in business and scientific domains. For this purpose, we present , our end-to-end trainable model using a two-step process for table parsing: Top Down, decomposing the table image into cells, and Bottom Up, associating cells to define the complete structure by assigning row and column spans. This method leverages constraints inspired by human cognition, such as cell alignment, continuity, and non-overlapping characteristics, optimizing the objective function to preserve these attributes. We introduce a vision transformer backbone augmented with local attention tailored for high-resolution input images. We also enhance inference speed for row and column adjacency prediction by replacing sampling-based graph neural networks with a single head self attention layer. Furthermore, we propose an effective split-and-merge strategy to handle highly dense tables without altering the network architecture. Additionally, we present problem-specific training strategies that improve performance across diverse visual characteristics and support cross-domain adaptations. Finally, through our robust protocol for comprehensive dataset analysis and result interpretation, we offer clear directions for future work in automated table structure recognition, highlighting its potential for substantial advancements in practical applications.

Abstract

Financial statements are critical for corporate analysis, fraud detection, and regulatory compliance, yet their complex tabular layouts, multi-year data structures, and hierarchical relationships pose significant challenges for automated reasoning. Existing Visual Question Answering (VQA) models struggle to deliver audit-ready responses due to their inability to explicitly localize evidence and reason over structured financial data. To bridge this gap, we introduce EviFiVQA, a large-scale benchmark dataset for evidence-based financial VQA, built on real-world financial statements. Leveraging FinTabNet, our dataset contains over 1.5 million question-answer pairs spanning five reasoning categories, each meticulously annotated with bounding boxes for relevant table cells to ensure traceable and interpretable responses. EviFiVQA introduces new challenges for Large Language Models (LLMs) and Vision-Language Models (VLMs) by requiring multi-step numerical reasoning, explicit evidence localization, and hierarchical financial aggregation—all within complex, irregular table layouts. We benchmark state-of-the-art models on EviFiVQA, revealing their limitations in evidence-grounded understanding of financial statements. We release EviFiVQA to drive progress in audit-ready financial AI, enhancing transparency and reliability in AI-driven financial analytics. The dataset and code are available at https://github.com/sachinraja13/EviFiVQA

Abstract

Airborne PM₁₀ poses significant respiratory and cardiovascular risks to the vulnerable population. In high-elevation urban zones like Gachibowli, Hyderabad, understanding the spatiotemporal behaviour of PM₁₀ spread is essential to develop targeted health advisories. This is not achievable solely by increasing the number of monitoring stations. The local topography, Land Use/Land Cover (LULC) patterns and mixing conditions must be understood to identify pollution sources and dispersal dynamics. We estimated PM₁₀ concentrations based on Aerosol Optical Depth from two satellite sources - half-hourly, INSAT-3D geostationary satellite data at 11km; and daily polar-orbiting, MODIS Aqua data at 3km spatial resolution. By validating these PM₁₀ estimates against government-monitored Central Pollution Control Board (CPCB) sensors and low-cost academic AirIOT sensors of IIIT-Hyderabad, a near-continuous and hyperlocal PM₁₀ concentration estimate, near the ground surface, has been generated. Among the 4 Machine learning models evaluated, Multiple Linear Regression technique offered the best results when temperature, humidity and ERA-5 Planetary Boundary Layer Height were used as predictors. While satellite-based PM₁₀ estimate averages ranged from 61–72 µg/m³, AirIOT and CPCB ranged 24–57 µg/m³ and 99–106 µg/m³ respectively, during colder months. While AirIOT sensors are spread across varied microenvironments, the CPCB node lies near a main road with partial green cover. Satellite-derived values provide frequent, broad-scale estimates of pollutant distribution, and AirIOT and CPCB give point measurements. This resolution gap drives our effort to bridge scales and produce PM₁₀ estimates at ~1–2 km resolution with temporal continuity, while also capturing sub-grid variability and dissipation patterns. Thus, this integrated framework allows for hyperlocal pollution mapping for health-related early warning systems even in areas with sparse PM₁₀ sensor coverage. By correlating PM₁₀ subtypes with LULC and dispersion, health outcomes can be traced to specific exposure types. By extending this method across India, to megacities like New Delhi, a health risk classification can be developed along with guidelines for clinical response measures.

Abstract

Handwritten Text Recognition (HTR) remains a challenging task, especially in Indic languages, due to varied handwriting styles, complex character structures, and the scarcity of annotated data. While deep learning models have shown promising results on curated benchmarks, their performance often degrades in real-world scenarios with high variability in writing. To address this, we propose a semi-supervised approach, SemiHastakshar, that leverages large-scale unlabeled word-level images collected from diverse sources across the Internet. We employ a high-confidence pseudo-labeling strategy to train on unlabeled samples iteratively. It allows the model to learn from a wider distribution of handwriting styles and improve generalization across writers. Our method demonstrates that combining a small amount of labeled data with a large unlabeled corpus leads to more robust HTR models for Indic scripts, advancing scalable recognition in low-resource settings. The code, model, and data will be publicly available for research.

Abstract

Abstract The stability of funicularstructuralformssuchasarches, vaults, and domes hasbeen a subject of fascinationinarchitectureengineeringefficientdueandtotheirstructuralperformance and aesthetics. Inthis context, methodologiesand procedures which drivetowards a reduction in carbonfootprint gain significance astheconstruction industry isresponsible for around 50% ofthe total carbon footprintglobally. A novel approach ispresentedtoidentifytheoptimal structural geometry ofan arch considering all criticalloadingscenarios.Akinematic framework basedon principles of limit analysisisutilised to identify thepossible collapse scenarios ofthearch.Keygeometricparameters such as span,riseto-span ratio, and thicknessare varied based on aMonteCarlo simulation to achievestructurally stable forms. ThedevelopedkinematicframeworkistrainedonMachineLearning (ML)models to predict the collapsecapacity. Based on this, anML-enabledoptimizationscheme is applied to thearches considering the criticalloading scenarios. Beyond itspracticalrelevance,theresearch revisits fundamentalquestionsaboutstructuralform, offering a thoughtfulintegration of mathematicalreasoning with architecturaland engineering practice.

Abstract

Modern large language models (LLMs) place extraordinary pres- sure on memory and compute budgets, making principled compres- sion indispensable for both deployment and continued training. We present Hierarchical Sparse Plus Low-Rank (HSS) compres- sion, a two–stage scheme that (i) removes the largest-magnitude weights into a sparse matrix 𝑆 and (ii) applies a recursive Hierar- chically Sparse Separable (HSS) low-rank factorisation to the dense residual matrix. A recursive rank-reducing strategy and a reverse Cuthill–Mckee (RCM) permutation are introduced to align high weights towards the diagonal with the block-diagonal hierarchy, maximising off-diagonal compressibility (because they are touched only once). HSS is hardware-friendly: its matrix–vector multiply reduces to one sparse and a sequence of thin-matrix multiplications, and can be trained end-to-end with standard optimisers. Experiments on LLaMA-7B show that targeting only the self- attention projections (1.6B parameters of 𝑄, 𝐾, and 𝑉 matrices out of a total 7B parameters) suffices to yield large memory savings while retaining comparable state-of-the-art perplexity scores on test samples of the WikiText dataset. For example, with a 30% sparsity budget and an outer rank of 512, sHSS-RCM achieves a perplexity of 1.64, outperforming dense baselines and classical sparse-plus-SVD variants, while also achieving significant memory savings.

Abstract

The lack of public data for Arabic-to-Hindi transliteration has hindered the develop- ment of systems that can handle the languages’ diverse linguistic styles. To address this, we introduce AH-Translit, a multi-domain dataset of 100K parallel pairs with over 1.2M Arabic and 1.5M Hindi words. We also present AH-Translit- Bench, a balanced, human-verified benchmark for fair evaluations across diverse linguistic domains. Our analysis reveals that domain-specific models, while strong in-domain, generalize poorly. We show that a single model, trained on a balanced mixture, achieves higher performance consistency across all domains. This ap- proach establishes a strong baseline with a Macro-averaged Character Error Rate (MaCER) of 15.7%. We release the benchmark and an evaluation package for reproducible, cross-domain assessment

Abstract

Large multilingual transformers such as XLM-RoBERTa achieve impressive performance on diverse NLP benchmarks, but understanding how they internally encode grammatical information remains challenging. This study investigates the encoding of syntactic and morphological information derived from the Paninian grammatical framework—specifically designed for morphologically rich Indian languages—across model layers. Using diagnostic probing, we analyze the hidden representations of frozen XLM-RoBERTa-base, mBERT, and IndicBERT models across seven Indian languages (Hindi, Kannada, Malayalam, Marathi, Telugu, Urdu, Bengali). Probes are trained to predict Paninian dependency relations (by edge probing) and essential morphosyntactic features (UPOS tags, Vibhakti markers). We find that syntactic structure (dependencies) is primarily encoded in the middle-to-upper-middle layers (layers 6–9), while lexical features peak slightly earlier. Although the general layer-wise trends are shared across models, significant variations in absolute probing performance reflect differences in model capacity, pre-training data, and language-specific characteristics. These findings shed light on how theory-specific grammatical information emerges implicitly within multilingual transformer representations trained largely on unstructured raw text.

Abstract

We introduce InterChart, a diagnostic benchmark that evaluates how well vision-language models (VLMs) reason across multiple related charts, a task central to real-world applications such as scientific reporting, financial analysis, and public policy dashboards. Unlike prior benchmarks focusing on isolated, visually uniform charts, InterChart challenges models with diverse question types ranging from entity inference and trend correlation to numerical estimation and abstract multi-step reasoning grounded in 2-3 thematically or structurally related charts. We organize the benchmark into three tiers of increasing difficulty: (1) factual reasoning over individual charts, (2) integrative analysis across synthetically aligned chart sets, and (3) semantic inference over visually complex, real-world chart pairs. Our evaluation of state-of-the-art open and closed-source VLMs reveals consistent and steep accuracy declines as chart complexity increases. We find that models perform better when we decompose multi-entity charts into simpler visual units, underscoring their struggles with cross-chart integration. By exposing these systematic limitations, InterChart provides a rigorous framework for advancing multimodal reasoning in complex, multi-visual environments.

Abstract

This paper presents a low-cost, Original Equipment Manufacturer (OEM) agnostic Internet of Things (IoT)-based system for monitoring the behaviour of Uninterruptible Power Supply (UPS) units during power outages and recovery. Frequent outages in developing regions cause equipment damage, operational downtime, and data loss. While UPS units provide backup power, affordable options for monitoring their performance remain limited. Commercial solutions such as Simple Network Management Protocol (SNMP) cards are expensive, manufacturer-specific, and reliant on network infrastructure, restricting their use in cost-sensitive or remote installations. The proposed system continuously records input and output currents to detect outages, switchovers, and UPS behaviour during these events, operating independently of mains power to ensure uninterrupted data capture. Deployed across 5 UPS installations, it collected over 4 million data points and automatically identified 61 outage events. Results demonstrate consistent and accurate detection with real-time visualisation via a web dashboard.