arXiv subjects

Kishor Datta Gupta

Publications and source records attributed to Kishor Datta Gupta.

At least 19 recordsLinked to original sources

arXiv2026

WALDO: One-Shot Exemplar-Conditioned Object Detection in Cluttered Scenes

Locating a specific object instance in a cluttered scene using a single reference image and a short description, and reporting when that instance is absent, large vision-language models usually address this task. We ask whether the same capability is available far more cheaply, from representations already learned by a world-model pretraining objective. We present WALDO, a one-shot exemplar- and language-conditioned detection head with 3.4M trainable parameters that reads frozen V-JEPA 2.1 features to jointly predict object localization and target presence, with no gradient on the backbone. Because exemplar-conditioned supervision is scarce, we synthesize training episodes from instance annotations, mining exemplars from ground-truth boxes and constructing absence cases that exclude the referenced instance while leaving same-category distractors in view. This is easy to get wrong: in the obvious implementation, crop size alone predicts the label, and a head trained on it reaches 0.9998 absence AUROC without ever consulting the exemplar, and we report the negative controls that close the shortcut. On 35 held-out cluttered scenes, WALDO achieves a 0.461 catalogue AP@50, compared to 0.306 for a prompted Grounding DINO baseline under an identical scorer. Substituting DINOv3 for V-JEPA under a matched 576-token grid drops within-category absence AUROC from 0.880 to 0.726 and instance AP@50 from 0.201 to 0.141, isolating the pretraining objective rather than input resolution as the source of the gain. Instance-level Success@1, however, reaches only 0.190 against a 0.190 category-chance floor: world-model features transfer to localization precision and absence detection but not to instance identity.

cs.CV↗

arXiv2026

TriCLE: Tri-Modal Vision-Language Reasoning for Edge-Deployed Fine-Grained Clustering

Edge platforms used for aerial observation must interpret aircraft imagery under limited memory, limited compute, and intermittent connectivity. This setting is difficult for standard RGB-only recognition models and general-purpose vision-language models, especially when calibrated thermal and LiDAR aircraft data are unavailable. We present TriCLE, an application-oriented tri-modal vision-language system for aircraft taxonomic grouping under edge constraints. From a single RGB aircraft image, TriCLE generates a structure-preserving FLIR-style thermal view and a pseudo-LiDAR depth projection, then fuses the aligned views with task instructions in a compact Qwen3-VL backbone. The model is aligned to an expert aircraft taxonomy based on propulsion, airframe family, size, design era, and configuration, so its outputs reflect engineering-relevant similarity rather than only surface appearance. We evaluate supervised fine-tuning, rotation-preserving SFT, and three policy-alignment strategies: GRPO, GSPO, and DAPO. Sequence-level GSPO gives the strongest validation performance, reaching 88.33\% validation accuracy and 0.91 weighted F1 on valid aircraft outputs. On a held-out aircraft test partition, GSPO achieves 78.00\% accuracy and 0.793 weighted F1 while preserving 94.00\% parseable output formatting. After 4-bit quantization and attention-memory optimization, the aligned 4B model fits an 8GB deployment target and processes each tri-modal triplet in 1.48 seconds. These results support TriCLE as a practical prototype for interpretable, edge-feasible aircraft grouping, while emphasizing the need for further validation on real aligned thermal and LiDAR sensor streams.

cs.CV↗

arXiv2026

GPF-LiveNews: A Streaming Evaluation Protocol for Group-Conditioned Framing in Large Language Models

Deployed language models are evaluated in a non-stationary environment: model versions, retrieval layers, safety systems, and real-world inputs all change over time. Static bias benchmarks remain useful, but they do not show how models frame newly emerging events for different prompted audiences. We introduce GPF-LIVENEWS, a streaming evaluation protocol and benchmark snapshot for auditing group-conditioned framing in open-ended LLM outputs. The protocol expands fresh BBC/Reuters news anchors across 42 identity labels and seven prompt families, then evaluates response bundles using semantic-sensitivity and sentiment-disparity signals. In a pilot over 12 monitoring runs and 23 hosted models, Policy/Action prompts produce the strongest semantic movement, while sentiment variation is flatter across dimensions and prompt families. The released artifact includes article metadata, prompt templates, instantiated prompts, model-output metadata, score tables, documentation, and reproduction scripts. We interpret all scores as observed-window audit signals for human review, not as permanent fairness rankings or direct proof of harmful bias.

cs.CL↗

arXiv2026

LiteVLA-H: Dual-Rate Vision-Language-Action Inference for Onboard Aerial Guidance and Semantic Perception

Vision-language-action (VLA) models have shown strong semantic grounding and task generalization in manipulation, but aerial deployment remains difficult because drones require low-latency closed-loop guidance under strict onboard compute and communication constraints. We present LiteVLA-H, a compact 256M-parameter VLA system designed for dual-rate operation on an NVIDIA Jetson AGX Orin: a fast outer-loop guidance mode for short action-token outputs and a slower semantic mode for scene understanding, hazard description, and operator-facing narration. The central empirical observation is that, in this compact edge regime, end-to-end latency is dominated by multimodal pre-fill rather than by the marginal cost of decoding a few extra tokens. This motivates a scheduler that issues reactive action tokens at 50.65,ms (19.74,Hz) while still supporting sentence-level semantic outputs at 149.90--164.57\ms (6.08--6.67,Hz) on the same embedded platform. To specialize the model without collapsing its descriptive competence, we use a knowledge-preserving fine-tuning recipe that mixes reactive flight data, aerial semantic data, and generic caption/VQA supervision. Beyond reporting current latency measurements, we position the system against recent state-of-the-art architectures, including AnywhereVLA, FutureVLA, and ReMem-VLA, showing that the measured action branch reaches a higher edge inference rate under our deployment conditions while retaining periodic semantic awareness.

cs.CV↗

arXiv2026

Physics-Based Benchmarking Metrics for Multimodal Synthetic Images

Current state of the art measures like BLEU, CIDEr, VQA score, SigLIP-2 and CLIPScore are often unable to capture semantic or structural accuracy, especially for domain-specific or context-dependent scenarios. For this, this paper proposes a Physics-Constrained Multimodal Data Evaluation (PCMDE) metric combining large language models with reasoning, knowledge based mapping and vision-language models to overcome these limitations. The architecture is comprised of three main stages: (1) feature extraction of spatial and semantic information with multimodal features through object detection and VLMs; (2) Confidence-Weighted Component Fusion for adaptive component-level validation; and (3) physics-guided reasoning using large language models for structural and relational constraints (e.g., alignment, position, consistency) enforcement.

cs.CV↗

arXiv2026

VLCE: A Knowledge-Enhanced Framework for Image Description in Disaster Assessment

General-purpose vision-language models (VLMs) such as LLaVA and QwenVL produce descriptions of disaster imagery that lack domain-specific vocabulary and actionable detail. We propose the Vision-Language Caption Enhancer (VLCE), a framework that integrates external semantic knowledge from ConceptNet and WordNet into the caption generation process for post-disaster satellite and UAV imagery. VLCE operates in two stages: first, a baseline VLM generates an initial caption conditioned on YOLOv8 object detections; second, a knowledge-enriched sequential model, a CNN-LSTM or a hierarchical cross-modal Transformer, refines the caption using a vocabulary augmented with 1,566 domain-relevant terms extracted from knowledge graphs. We evaluate VLCE on two disaster benchmarks: xBD (satellite, 6,369 images, 3 damage classes) and RescueNet (UAV, 4,494 images, 12 damage classes), using CLIPScore for semantic alignment and InfoMetIC for informativeness. On RescueNet with the Transformer decoder, VLCE with knowledge graph enrichment produces captions preferred over QwenVL baselines in 95.33% of image pairs on InfoMetIC and 73.64% on CLIPScore. Qualitative analysis shows that without knowledge graph integration, generated captions exhibit hallucinations, word repetition, and semantic incoherence, whereas knowledge-enriched captions maintain factual consistency and domain-appropriate vocabulary.

cs.CV↗

arXiv2026

LiteVLA-Edge: Quantized On-Device Multimodal Control for Embedded Robotics

Vision-Language-Action (VLA) models provide a unified framework for perception, language conditioning, and action generation, but many existing systems remain difficult to deploy in embedded robotic settings because of their computational requirements and inference latency. In this paper, we present LiteVLA-Edge, a deployment-oriented VLA pipeline for fully on-device inference on Jetson Orin-class hardware. Our approach combines supervised image-to-action fine-tuning in FP32 with post-training 4-bit GGUF quantization and GPU-accelerated inference through the \texttt{llama.cpp} runtime. Under our deployment configuration, LiteVLA-Edge achieves a mean end-to-end latency of 150.5\,ms (approximately 6.6\,Hz) while operating entirely offline within a ROS~2-integrated perception--reasoning--action pipeline. Rather than introducing a new policy objective, our contribution is a practical systems path for executing compact multimodal control models locally on embedded hardware while preserving modular interfaces between perception, reasoning, and actuation. These results establish timing feasibility for reactive language-conditioned control and provide a reproducible baseline for future task-level evaluation of on-device VLAs in robotics.

cs.RO↗

arXiv2026

Verbatim Data Transcription Failures in LLM Code Generation: A State-Tracking Stress Test

Many real-world software tasks require exact transcription of provided data into code, such as cryptographic constants, protocol test vectors, allowlists, and calibration tables. These tasks are operationally sensitive because small omissions or alterations can remain silent while producing syntactically valid programs. This paper introduces a deliberately minimal transcription-to-code benchmark to isolate this reliability concern in LLM-based code generation. Given a list of high-precision decimal constants, a model must generate Python code that embeds the constants verbatim and performs a simple aggregate computation. We describe the prompting variants, evaluation protocol based on exact-string inclusion, and analysis framework used to characterize state-tracking and long-horizon generation failures. The benchmark is intended as a compact stress test that complements existing code-generation evaluations by focusing on data integrity rather than algorithmic reasoning.

cs.SE↗

arXiv2025

TinyLLM: Evaluation and Optimization of Small Language Models for Agentic Tasks on Edge Devices

This paper investigates the effectiveness of small language models (SLMs) for agentic tasks (function/tool/API calling) with a focus on running agents on edge devices without reliance on cloud infrastructure. We evaluate SLMs using the Berkeley Function Calling Leaderboard (BFCL) framework and describe parameter-driven optimization strategies that include supervised fine-tuning (SFT), parameter-efficient fine-tuning (PEFT), reinforcement learning (RL)-based optimization, preference alignment via Direct Preference Optimization (DPO), and hybrid methods. We report results for models including TinyAgent, TinyLlama, Qwen, and xLAM across BFCL categories (simple, multiple, parallel, parallel-multiple, and relevance detection), both in live and non-live settings, and in multi-turn evaluations. We additionally detail a DPO training pipeline constructed from AgentBank data (e.g., ALFRED), including our conversion of SFT data to chosen-rejected pairs using TinyLlama responses as rejected outputs and manual validation. Our results demonstrate clear accuracy differences across model scales where medium-sized models (1-3B parameters) significantly outperform ultra-compact models (<1B parameters), achieving up to 65.74% overall accuracy, and 55.62% multi-turn accuracy with hybrid optimization. This study highlights the importance of hybrid optimization strategies that enable small language models to deliver accurate, efficient, and stable agentic AI on edge devices, making privacy-preserving, low-latency autonomous agents practical beyond the cloud.

cs.LG↗

arXiv2025

Lite VLA: Efficient Vision-Language-Action Control on CPU-Bound Edge Robots

The deployment of artificial intelligence models at the edge is increasingly critical for autonomous robots operating in GPS-denied environments where local, resource-efficient reasoning is essential. This work demonstrates the feasibility of deploying small Vision-Language Models (VLMs) on mobile robots to achieve real-time scene understanding and reasoning under strict computational constraints. Unlike prior approaches that separate perception from mobility, the proposed framework enables simultaneous movement and reasoning in dynamic environments using only on-board hardware. The system integrates a compact VLM with multimodal perception to perform contextual interpretation directly on embedded hardware, eliminating reliance on cloud connectivity. Experimental validation highlights the balance between computational efficiency, task accuracy, and system responsiveness. Implementation on a mobile robot confirms one of the first successful deployments of small VLMs for concurrent reasoning and mobility at the edge. This work establishes a foundation for scalable, assured autonomy in applications such as service robotics, disaster response, and defense operations.

cs.RO↗

arXiv2025

Beyond Visual Similarity: Rule-Guided Multimodal Clustering with explicit domain rules

Traditional clustering techniques often rely solely on similarity in the input data, limiting their ability to capture structural or semantic constraints that are critical in many domains. We introduce the Domain Aware Rule Triggered Variational Autoencoder (DARTVAE), a rule guided multimodal clustering framework that incorporates domain specific constraints directly into the representation learning process. DARTVAE extends the VAE architecture by embedding explicit rules, semantic representations, and data driven features into a unified latent space, while enforcing constraint compliance through rule consistency and violation penalties in the loss function. Unlike conventional clustering methods that rely only on visual similarity or apply rules as post hoc filters, DARTVAE treats rules as first class learning signals. The rules are generated by LLMs, structured into knowledge graphs, and enforced through a loss function combining reconstruction, KL divergence, consistency, and violation penalties. Experiments on aircraft and automotive datasets demonstrate that rule guided clustering produces more operationally meaningful and interpretable clusters for example, isolating UAVs, unifying stealth aircraft, or separating SUVs from sedans while improving traditional clustering metrics. However, the framework faces challenges: LLM generated rules may hallucinate or conflict, excessive rules risk overfitting, and scaling to complex domains increases computational and consistency difficulties. By combining rule encodings with learned representations, DARTVAE achieves more meaningful and consistent clustering outcomes than purely data driven models, highlighting the utility of constraint guided multimodal clustering for complex, knowledge intensive settings.

cs.LG↗

arXiv2025

Continuous Monitoring of Large-Scale Generative AI via Deterministic Knowledge Graph Structures

Generative AI (GEN AI) models have revolutionized diverse application domains but present substantial challenges due to reliability concerns, including hallucinations, semantic drift, and inherent biases. These models typically operate as black-boxes, complicating transparent and objective evaluation. Current evaluation methods primarily depend on subjective human assessment, limiting scalability, transparency, and effectiveness. This research proposes a systematic methodology using deterministic and Large Language Model (LLM)-generated Knowledge Graphs (KGs) to continuously monitor and evaluate GEN AI reliability. We construct two parallel KGs: (i) a deterministic KG built using explicit rule-based methods, predefined ontologies, domain-specific dictionaries, and structured entity-relation extraction rules, and (ii) an LLM-generated KG dynamically derived from real-time textual data streams such as live news articles. Utilizing real-time news streams ensures authenticity, mitigates biases from repetitive training, and prevents adaptive LLMs from bypassing predefined benchmarks through feedback memorization. To quantify structural deviations and semantic discrepancies, we employ several established KG metrics, including Instantiated Class Ratio (ICR), Instantiated Property Ratio (IPR), and Class Instantiation (CI). An automated real-time monitoring framework continuously computes deviations between deterministic and LLM-generated KGs. By establishing dynamic anomaly thresholds based on historical structural metric distributions, our method proactively identifies and flags significant deviations, thus promptly detecting semantic anomalies or hallucinations. This structured, metric-driven comparison between deterministic and dynamically generated KGs delivers a robust and scalable evaluation framework.

cs.AI↗

arXiv2025

UrbanInsight: A Distributed Edge Computing Framework with LLM-Powered Data Filtering for Smart City Digital Twins

Cities today generate enormous streams of data from sensors, cameras, and connected infrastructure. While this information offers unprecedented opportunities to improve urban life, most existing systems struggle with scale, latency, and fragmented insights. This work introduces a framework that blends physics-informed machine learning, multimodal data fusion, and knowledge graph representation with adaptive, rule-based intelligence powered by large language models (LLMs). Physics-informed methods ground learning in real-world constraints, ensuring predictions remain meaningful and consistent with physical dynamics. Knowledge graphs act as the semantic backbone, integrating heterogeneous sensor data into a connected, queryable structure. At the edge, LLMs generate context-aware rules that adapt filtering and decision-making in real time, enabling efficient operation even under constrained resources. Together, these elements form a foundation for digital twin systems that go beyond passive monitoring to provide actionable insights. By uniting physics-based reasoning, semantic data fusion, and adaptive rule generation, this approach opens new possibilities for creating responsive, trustworthy, and sustainable smart infrastructures.

cs.AI↗

arXiv2025

Principles and Components of Federated Learning Architectures

Federated Learning (FL) is a machine learning framework where multiple clients, from mobiles to enterprises, collaboratively construct a model under the orchestration of a central server but still retain the decentralized nature of the training data. This decentralized training of models offers numerous advantages, including cost savings, enhanced privacy, improved security, and compliance with legal requirements. However, for all its apparent advantages, FL is not immune to the limitations of conventional machine learning methodologies. This article provides an elaborate explanation of the inherent concepts and features found within federated learning architecture, addressing five key domains: system heterogeneity, data partitioning, machine learning models, communication protocols, and privacy techniques. This article also highlights the limitations in this domain and proposes avenues for future work. Besides, we provide a set of architectural patterns for federated learning systems, which are derived from the systematic survey of the literature. The main elements of FL, the fundamentals of Federated Learning, and a few architectural specifics will all be better understood with the aid of this research.

cs.LG↗

arXiv2025

Trustworthy XAI and Application

Artificial Intelligence (AI) is an important part of our everyday lives. We use it in self-driving cars and smartphone assistants. People often call it a "black box" because its complex systems, especially deep neural networks, are hard to understand. This complexity raises concerns about accountability, bias, and fairness, even though AI can be quite accurate. Explainable Artificial Intelligence (XAI) is important for building trust. It helps ensure that AI systems work reliably and ethically. This article looks at XAI and its three main parts: transparency, explainability, and trustworthiness. We will discuss why these components matter in real-life situations. We will also review recent studies that show how XAI is used in different fields. Ultimately, gaining trust in AI systems is crucial for their successful use in society.

cs.AI↗

arXiv2025

Advanced Tool Learning and Selection System (ATLASS): A Closed-Loop Framework Using LLM

The combination of LLM agents with external tools enables models to solve complex tasks beyond their knowledge base. Human-designed tools are inflexible and restricted to solutions within the scope of pre-existing tools created by experts. To address this problem, we propose ATLASS, an advanced tool learning and selection system designed as a closed-loop framework. It enables the LLM to solve problems by dynamically generating external tools on demand. In this framework, agents play a crucial role in orchestrating tool selection, execution, and refinement, ensuring adaptive problem-solving capabilities. The operation of ATLASS follows three phases: The first phase, Understanding Tool Requirements, involves the Agents determining whether tools are required and specifying their functionality; the second phase, Tool Retrieval/Generation, involves the Agents retrieving or generating tools based on their availability; and the third phase, Task Solving, involves combining all the component tools necessary to complete the initial task. The Tool Dataset stores the generated tools, ensuring reusability and minimizing inference cost. Current LLM-based tool generation systems have difficulty creating complex tools that need APIs or external packages. In ATLASS, we solve the problem by automatically setting up the environment, fetching relevant API documentation online, and using a Python interpreter to create a reliable, versatile tool that works in a wider range of situations. OpenAI GPT-4.0 is used as the LLM agent, and safety and ethical concerns are handled through human feedback before executing generated code. By addressing the limitations of predefined toolsets and enhancing adaptability, ATLASS serves as a real-world solution that empowers users with dynamically generated tools for complex problem-solving.

cs.AI↗

arXiv2025

A Comprehensive Review on Understanding the Decentralized and Collaborative Approach in Machine Learning

The arrival of Machine Learning (ML) completely changed how we can unlock valuable information from data. Traditional methods, where everything was stored in one place, had big problems with keeping information private, handling large amounts of data, and avoiding unfair advantages. Machine Learning has become a powerful tool that uses Artificial Intelligence (AI) to overcome these challenges. We started by learning the basics of Machine Learning, including the different types like supervised, unsupervised, and reinforcement learning. We also explored the important steps involved, such as preparing the data, choosing the right model, training it, and then checking its performance. Next, we examined some key challenges in Machine Learning, such as models learning too much from specific examples (overfitting), not learning enough (underfitting), and reflecting biases in the data used. Moving beyond centralized systems, we looked at decentralized Machine Learning and its benefits, like keeping data private, getting answers faster, and using a wider variety of data sources. We then focused on a specific type called federated learning, where models are trained without directly sharing sensitive information. Real-world examples from healthcare and finance were used to show how collaborative Machine Learning can solve important problems while still protecting information security. Finally, we discussed challenges like communication efficiency, dealing with different types of data, and security. We also explored using a Zero Trust framework, which provides an extra layer of protection for collaborative Machine Learning systems. This approach is paving the way for a bright future for this groundbreaking technology.

cs.LG↗

arXiv2025

SOK: Exploring Hallucinations and Security Risks in AI-Assisted Software Development with Insights for LLM Deployment

The integration of Large Language Models (LLMs) such as GitHub Copilot, ChatGPT, Cursor AI, and Codeium AI into software development has revolutionized the coding landscape, offering significant productivity gains, automation, and enhanced debugging capabilities. These tools have proven invaluable for generating code snippets, refactoring existing code, and providing real-time support to developers. However, their widespread adoption also presents notable challenges, particularly in terms of security vulnerabilities, code quality, and ethical concerns. This paper provides a comprehensive analysis of the benefits and risks associated with AI-powered coding tools, drawing on user feedback, security analyses, and practical use cases. We explore the potential for these tools to replicate insecure coding practices, introduce biases, and generate incorrect or non-sensical code (hallucinations). In addition, we discuss the risks of data leaks, intellectual property violations and the need for robust security measures to mitigate these threats. By comparing the features and performance of these tools, we aim to guide developers in making informed decisions about their use, ensuring that the benefits of AI-assisted coding are maximized while minimizing associated risks.

cs.SE↗