arXiv subjects

M. F. Mridha

Publications and source records attributed to M. F. Mridha.

At least 19 recordsLinked to original sources

arXiv2026

LENS-GRF: Permutation-Invariant Lesion Evidence Network with Gated Residual Fusion for Acne Severity Grading and Multi-Rater Clinical Oracle Analysis

Localized lesion evidence may improve automated acne severity grading, but detector errors and differences in clinical grading criteria can limit generalization. We evaluated LENS-GRF, a lesion evidence network with permutation-invariant aggregation and gated residual fusion that combines a global Vision Transformer with lesion-region embeddings derived from YOLO11s proposals. Fully automated four-class grading was evaluated on 292 held-out ACNE04 images and, without external adaptation, on 200 PLSBRACNE01 participants represented by 600 facial views. Annotation-assisted inference used registered lesion boxes from three dermatologists in a matched subset of 148 participants. Performance measures included accuracy, macro-averaged precision, sensitivity, and one-versus-rest specificity. Paired comparisons used exact McNemar tests with Holm adjustment. On ACNE04, automated LENS-GRF achieved 80.82% accuracy, 78.80% macro precision, 78.92% sensitivity, and 92.78% specificity. With supplied lesion boxes, internal accuracy averaged 95.89% across three seeds. On the full external cohort, automated LENS-GRF achieved 35.00% accuracy, compared with 42.50% for an independently trained global baseline. External macro precision, sensitivity, and specificity for LENS-GRF were 50.52%, 34.31%, and 78.19%, respectively. In the matched subset, accuracy using Doctor A's lesion annotations was 47.97%, compared with 35.14% for automated inference (unadjusted $p = 0.0094$; Holm-adjusted $p = 0.0846$). Strong internal performance with supplied lesion annotations did not translate into reliable automated grading on the external cohort. More reliable lesion detection, modelling that accounts for differences in grading criteria, and independent validation are needed before clinical use.

cs.CV↗

arXiv2026

How Reliable Are Object Detection Models Beyond Clear Weather? A Deployment-Oriented Robustness Study for Autonomous Driving

Autonomous driving perception systems must maintain reliable object detection under diverse environmental conditions, not only in clear daytime scenes. However, the cross-weather robustness of modern detectors remains insufficiently characterised under source-only evaluation. This study evaluates representative object detection architectures trained exclusively on clear daytime driving scenes and tested without adaptation under unseen fog, rain, snow, and night conditions. The evaluation covers lightweight and medium-scale one-stage, two-stage, and transformer-based detectors within a unified protocol. Robustness is assessed using performance Drop (%), a Weather Robustness Index (WRI), and class-level sensitivity analysis for critical road participants, including cars, buses, trucks, and persons. Prediction confidence reliability under environmental domain shift is measured using Expected Calibration Error (ECE). Results reveal consistent performance degradation in adverse conditions, reaching 62.90% at night. Buses show the highest sensitivity to environmental variability, while cars remain comparatively stable across weather transitions. Comparisons indicate that YOLO-based detectors exhibit the lowest calibration mismatch, reflecting stronger agreement between predicted confidence and empirical detection correctness. In contrast, DETR achieves the lowest robustness score, with a WRI of 0.6859, indicating greater vulnerability to cross-weather domain shift. These findings offer deployment-oriented insights into detector robustness and confidence reliability for safety-critical intelligent vehicle perception systems.

cs.CV↗

arXiv2026

CRaFT-XDet: Causal and Robust Fusion Transformer for Dual-Energy X-ray Prohibited Item Detection

Automated prohibited item detection in X-ray security imagery is fundamentally constrained by severe object overlap, scanner-dependent domain shifts, and spurious correlations induced by cluttered baggage composition. Existing detectors largely rely on appearance-driven supervision, limiting their robustness and cross-scanner generalisation. We introduce CRaFT-XDet, a causality-grounded and physics- motivated detection framework that reformulates X-ray threat detection as invariant representation learning under attenuation-driven modality variation. Our approach is grounded in a Scanner-Invariant Causal Detection formalisation (Theorem 1), which establishes that minimising the proposed Causal Invariance Regularisation (CIR) objective across partitioned scanner environments approximates the interventional distribution P (Y | do(Xobj)) under standard identifiability assumptions, providing a principled causal formulation for dual-energy X-ray detection. CRaFT-XDet integrates: (i) causal feature disentanglement to suppress context-induced shortcuts; (ii) physics-motivated regularisation to enforce material-consistent embeddings; and (iii) transformer-based cross-energy interaction modelling for structured reasoning in cluttered scenes. With 29.2 M parameters, 89.5 GFLOPs, and 345 FPS (1.7 ms inference), CRaFT-XDet is real-time deployable. Trained on SIXray with five-fold cross-validation, the framework achieves 0.9389 mAP@0.5 and 0.8021 mAP@0.5:0.95. Without domain adaptation, CRaFT-XDet maintains strong cross- dataset performance on OPIXray and HiXray, indicating robust generalization across heterogeneous scanner distributions. In category-agnostic binary threat detection, the model attains 93.92% recall and 0.9031 mean IoU, highlighting reliability in safety-critical screening where missed detections are costly.

cs.CV↗

arXiv2026

Certified Mechanistic Edits: Behavioral Guarantees for Skill Removal and Preservation

Mechanistic edits (ablations, weight edits, activation steering) are the standard tools for unlearning a harmful capability from a neural network while preserving useful ones. Current approaches validate their effects only by testing, which can never cover an entire continuous region of inputs. Prior work at the interpretability-verification boundary certifies descriptions of a model: what a circuit computes, or whether it faithfully explains the whole. We instead certify the behavioral effect of an edit: that disabling a circuit removes one skill and provably preserves another, for every input in a region; a feature non-interference guarantee in the information-flow-security sense. We demonstrate such certified edits from toy ReLU networks up to a standard softmax + LayerNorm transformer, proving removal and preservation over continuous embedding-space regions and reaching roughly 9x the input-perturbation dimension an exact solver can handle by switching to sound bound propagation. Furthermore, we prove that no finite deterministic black-box test can certify removal, exhibiting an edit that passes exhaustive testing yet provably fails on a survivor pocket that can be made arbitrarily small. Guarantees hold on small, standard-architecture networks and, like any removal claim, presuppose that the target skill admits a decidable specification, a property which real-world harms may not have.

cs.LG↗

arXiv2026

AI in Supply Chain Risk Assessment: A Systematic Literature Review and Bibliometric Analysis

Supply chain risk assessment (SCRA) is pivotal for ensuring resilience in increasingly complex global supply networks. While existing reviews have explored traditional methodologies, they often neglect emerging artificial intelligence (AI) and machine learning (ML) applications and mostly lack combined systematic and bibliometric analyses. This study addresses these gaps by integrating a systematic literature review with bibliometric analysis, examining 1,903 articles (2015-2025), with 54 studies selected through PRISMA guidelines. Our findings reveal that ML models, including Random Forest, XGBoost, and hybrid approaches, significantly improve risk prediction accuracy and adaptability in post-pandemic contexts. The bibliometric analysis identifies key trends, influential authors, and institutional contributions, highlighting China and the United States as leading research hubs. Practical insights emphasize the integration of explainable AI (XAI) for transparent decision-making, real-time data utilization, and blockchain for traceability. The study underscores the necessity of dynamic strategies, interdisciplinary collaboration, and continuous model evaluation to address challenges such as data quality and interpretability. By synthesizing AI-driven methodologies with resilience frameworks, this review provides actionable guidance for optimizing supply chain risk management, fostering adaptability, and informing future research in evolving risk landscapes.

cs.LG↗

arXiv2026

NetInjectBench: Benchmarking Indirect Prompt Injection in Tool-Using Large Language Model Agents for Network Operations

Tool-using large language model (LLM) agents are attractive for network operations, but tickets, alerts, logs, runbooks, and ChatOps messages can carry indirect prompt injections. We present NetInjectBench, a 130-scenario benchmark that separates untrusted artifact text, trusted policy metadata, and evaluation labels for network-operation tool use. The sample contains 40 benign, 40 weak-attack, 40 strong-attack, and 10 approved high-impact change scenarios; each is evaluated with Qwen2.5-7B, Llama3.1-8B, and Mistral-7B. Across 240 attack instances, naive execution reached an 82.50% unsafe tool-action rate. Prompt-only safety, Self-Reminder, Spotlighting, and a Two-Pass LLM Judge reduced this rate to 25.63%, 21.67%, 18.33%, and 10.00%, respectively. Static allowlisting reached 5.00% but blocked all approved changes, yielding 0.00% usefulness and 100.00% overblocking on approved cases. Under the stated metadata-integrity assumption, the metadata-aware policy gate produced 0/240 unsafe attack actions, with a 95% Wilson upper bound of 1.58%, while preserving 99.17% attack-scenario usefulness and 100.00% approved-change usefulness. The findings show that network-operation agents need execution-time authorization boundaries alongside prompt-level instruction hygiene.

cs.CR↗

arXiv2026

SkinAgent AI: A Safety-Grounded Multimodal Agentic Framework for Non-Diagnostic Skincare Support

Consumer-facing skincare AI must coordinate visual evidence, product information, tool use, and user-facing actions within explicit evidence and safety boundaries. This study evaluates SkinAgent AI, a non-diagnostic multimodal framework that combines visual concern routing with grounded and auditable LLM-based orchestration. The architecture includes routing for Acne, Pores, and Wrinkles; photograph-based skin-type estimation; count-informed ordinal acne-severity support; typed tools; database-grounded recommendation and action functions; deterministic safety, privacy, and evidence checks; approval before state-changing actions; and structured trace and replay mechanisms. Visual-model performance and system-level agent behavior were evaluated separately. Across three seeds, the skin-condition routing model achieved 99.84% +/- 0.07% accuracy. Skin-type estimation achieved 88.85% accuracy, while count-informed acne-severity support achieved 84.59% accuracy with a quadratic weighted kappa of 0.9076. On a locked but non-independent 240-case system benchmark, intent accuracy was 80.00%, exact tool-set match was 62.92%, and strict task completion was 47.08%. No violations or successful cross-user leakage events were observed in the finite safety and privacy test suites. Tool-selection errors, incomplete grounding of product attributes, and unreliable failure fallback nevertheless remained. These findings support the feasibility of bounded, database-grounded, and traceable agent orchestration for non-diagnostic skincare assistance. They do not establish clinical readiness, external generalization, formal privacy guarantees, or universal safety. Independent validation, expert assessment, robustness and fairness testing, and prospective evaluation in real-world settings remain necessary.

cs.AI↗

arXiv2026

CLAIR-Fin: An Adversarial Multi-Agent Framework for Claim-Level Verification and Adaptive Debate in Cross-Modal Financial QA

Existing defenses against hallucination in retrieval-augmented and multi-agent pipelines remain partial: evidence is trusted despite modality disagreement, debate verifies an aggregate report rather than individual claims, and such verification occurs only after drafting, leaving inter-agent errors undetected until the final text. To close this gap, we present CLAIR-Fin, a nine-agent framework that decomposes each question into atomic claims maintained in a typed Financial Claim Ledger. Each claim is resolved through Asymmetric Evidence Authority, which conditions evidence trust on claim type rather than treating all modalities as equally reliable; Chain-of-Custody Verification, which checks grounding at the hand-off between drafting and adversarial review rather than only at the pipeline's exit; an Adaptive Rebuttal Cycle, which routes contested claims through adversarial debate whose depth scales with what that debate finds; and a terminal entailment audit paired with a continuous Hallucination Risk Index that distinguishes claims that passed scrutiny from claims never contested. We evaluate CLAIR-Fin on BB-FinQA-X, a 500-question cross-modal financial evaluation set built from Bangladesh Bank Annual Report material, stratified by query type, format, and difficulty. Relative to a single-pass retrieval-augmented generation baseline, it raises faithfulness ($0.780 \rightarrow 0.889$) while abstaining on 5.4% of questions when evidence is insufficient rather than forcing an unsupported response, and it exceeds stronger retrieval-strategy baselines such as HyDE and Graph-RAG on faithfulness ($\leq 0.874$).

cs.CL↗

arXiv2026

EduGuard: A Safe RAG-Based LLM Tutor for Programming Education

Generative AI (GenAI) is increasingly used by students for programming explanation, debugging, and assignment support. Yet unrestricted large language model (LLM) tutors can hallucinate, contradict course policy, reveal complete solutions, and foster passive dependence. This paper presents EduGuard, a safe retrieval-augmented generation (RAG) tutoring framework for introductory programming. EduGuard integrates query understanding, instructor-approved course retrieval, pedagogical strategy selection, rubric-aware generation, claim-level verification, and overreliance control. To make evaluation provenance explicit, we construct BILearn-CS, a 600-query instructor-authored, TA-validated benchmark spanning concept questions, debugging cases, misconceptions, assignment-support requests, code-mixed Bangla-English queries, and adversarial direct-answer prompts. Moving beyond a synthetic-only benchmark, we further evaluate on a 150-query public CS50-style course-forum set and run a small controlled pilot with 10 undergraduates using a counterbalanced pre-test/post-test design. Using Meta-Llama-3.1-8B-Instruct as the primary generator, hybrid FAISS/BM25 retrieval, and DeBERTa-v3-large-MNLI as an architecturally separate verifier, EduGuard is compared against strong baselines: GPT-4o-mini Tutor, Llama Socratic Tutor, LPITutor-style RAG, RAG with rubric prompting, and RAG with same-model self-checking. On BILearn-CS, EduGuard attains the best correctness (90.1%), grounding (89.4%), and rubric alignment (90.8%), with the lowest hallucination (4.9%) and direct-answer leakage (9.8%). In the pilot, it raises immediate post-test accuracy from 68.4% to 81.2% and cuts overreliance from 38.0% to 17.0% relative to GPT-4o-mini Tutor. These results suggest safe GenAI tutoring requires not only retrieval or strong prompting, but explicit pedagogical control, evidence verification, and deployment safeguards.

cs.CY↗

arXiv2026

Designing Cellular Manufacturing Systems in the Presence of Alternative Process Plans

In the design of cellular manufacturing systems (CMS), numerous technological and managerial decisions must be made at both the design and operational stages. The first step in designing a CMS involves grouping parts and machines. In this paper, four integer programming formulations are presented for grouping parts and machines in a CMS at both the design and operational levels for a generalised grouping problem, where each part has more than one process plan and each operation of a process plan can be performed on more than one machine. These four integer programming formulations have the ability to simultaneously optimise cell formation and operational routing decisions within a unified framework, whereas prior models often treat these as sequential or independent stages. By integrating machine flexibility and alternative process plans directly into the mathematical objective, this approach ensures a more globally optimal configuration for generalised grouping problems. Minimising inter-cell and intra-cell movements is achieved by assigning as many consecutive operations of a part type as possible to the same cell and machine. This objective is evaluated against alternatives like reducing machine investment and operating costs. Numerical examples demonstrate how the proposed formulations work and highlight their effectiveness in practice.

cs.AI↗

arXiv2026

CardioMeta: Calibrated Multi-Task Prediction of Diabetes, Hypertension, and Cardiovascular Disease Across Population and EHR Data

Cardiometabolic diseases remain among the most persistent drivers of preventable morbidity because diabetes, hypertension, and cardiovascular disease frequently co-occur and share metabolic, vascular, demographic, and behavioral determinants. Existing machine learning studies for chronic disease prediction often emphasize discrimination on a single dataset, while underreporting label leakage, calibration, temporal robustness, external transportability, and subgroup reliability. This paper presents CardioMeta, a calibrated multi-task framework for joint prediction of diabetes, hypertension, and cardiovascular disease across population survey and electronic health record (EHR) data. The study uses NHANES for population-level model development and temporal validation, and MIMIC-IV for EHR-domain evaluation under substantial distribution shift. To reduce circular label reconstruction, the primary analysis excludes disease-defining variables from the corresponding prediction heads, while a full-clinical feature setting is retained only as sensitivity analysis. CardioMeta combines a shared cardiometabolic encoder with disease-specific gated heads and post-hoc probability calibration. In the leakage-reduced temporal validation setting, the model achieved a macro-AUROC of 0.839, macro-AUPRC of 0.536, macro-F1 of 0.614, and expected calibration error of 0.024, with modest but consistent improvements over strong gradient-boosting and neural tabular baselines. External evaluation on MIMIC-IV showed clear degradation under domain shift, while limited fine-tuning partially recovered performance. The findings indicate that the principal value of multi-task cardiometabolic modeling lies not in inflated accuracy, but in reproducible leakage control, calibrated probabilities, and transparent reliability reporting across heterogeneous healthcare data sources.

cs.LG↗

arXiv2026

LabourCrew: A Multi-Agent RAG Framework for Trustworthy Adversarial Deliberation and Statutory Reasoning over Labour Law

In statutory question answering, every claim must be traceable to evidence, not merely relevant, since unverifiable labour-rights answers carry serious legal consequences. Current systems fall short: single-pass RAG cannot detect insufficient evidence, while multi-agent legal-debate systems treat grounding as a prompting convention, letting agents cite unretrieved evidence. To address this gap, we introduce LabourCrew, a multi-agent RAG framework built around three grounding mechanisms: StatuteGraph, a graph index that explicitly links chapter, section, proviso, and cross-reference structure rather than fixed-length spans; an Evidence Exchange Protocol that confines advocates and an interpreter to an evidence ledger, making citation to unretrieved text impossible, while a fault-tolerant supervisor board runs advocates in parallel so individual failures degrade rather than crash the system; and a Calibrated Trust Gate that replaces categorical accept/reject decisions with a trust score, thresholded via conformal risk control for a distribution-free bound on the false-accept rate. We evaluate on LabourActQA, a 500-item Bangla question set from the Bangladesh Labour Act, 2006, spanning seven reasoning categories and three difficulty tiers. The framework drives the empirical false-accept rate to 0.081, within the target level ($α= 0.10$), achieves the highest Answer Relevancy among HyDE RAG, Graph-RAG, and Hierarchical RAG (0.862 $>$ 0.839, 0.815, 0.828), and degrades gradually rather than catastrophically as question difficulty increases. These results show that calibrated abstention, not retrieval quality alone, is what makes legal question answering auditable in low-resource statutory domains.

cs.CL↗

arXiv2026

Vision Transformers for End-to-End Quark-Gluon Jet Classification from Calorimeter Images

Distinguishing between quark- and gluon-initiated jets is a critical and challenging task in high-energy physics, pivotal for improving new physics searches and precision measurements at the Large Hadron Collider. While deep learning, particularly Convolutional Neural Networks (CNNs), has advanced jet tagging using image-based representations, the potential of Vision Transformer (ViT) architectures, renowned for modeling global contextual information, remains largely underexplored for direct calorimeter image analysis, especially under realistic detector and pileup conditions. This paper presents a systematic evaluation of ViTs and ViT-CNN hybrid models for quark-gluon jet classification using simulated 2012 CMS Open Data. We construct multi-channel jet-view images from detector-level energy deposits (ECAL, HCAL) and reconstructed tracks, enabling an end-to-end learning approach. Our comprehensive benchmarking demonstrates that ViT-based models, notably ViT+MaxViT and ViT+ConvNeXt hybrids, consistently outperform established CNN baselines in F1-score, ROC-AUC, and accuracy, highlighting the advantage of capturing long-range spatial correlations within jet substructure. This work establishes the first systematic framework and robust performance baselines for applying ViT architectures to calorimeter image-based jet classification using public collider data, alongside a structured dataset suitable for further deep learning research in this domain.

cs.CV↗

arXiv2026

Optimizing Container Loading and Unloading through Dual-Cycling and Dockyard Rehandle Reduction Using a Hybrid Genetic Algorithm

This paper addresses the NP-hard problem of optimizing container handling at ports by integrating Quay Crane Dual-Cycling (QCDC) and dockyard rehandle minimization. We realized that there are interdependencies between the unloading sequence of QCDC and the dockyard plan and propose the Quay Crane Dual Cycle - Dockyard Rehandle Genetic Algorithm (QCDC-DR-GA), a hybrid Genetic Algorithm (GA) that holistically optimizes both aspects: maximizing the number of Dual Cycles (DCs) and minimizing the number of dockyard rehandles. QCDC-DR-GA employs specialized crossover and mutation strategies. Extensive experiments on various ship sizes demonstrate that QCDC-DR-GA reduces total operation time by 15-20% for large ships compared to existing methods. Statistical validation via two-tailed paired t-tests confirms significant improvements at a 5% significance level. The results underscore the inefficiency of isolated optimization and highlight the critical need for integrated algorithms in port operations. This approach increases resource utilization and operational efficiency, offering a cost-effective solution for ports to decrease turnaround times without infrastructure investments.

cs.NE↗

arXiv2026

Predicting Male Domestic Violence Using Explainable Ensemble Learning and Exploratory Data Analysis

Domestic violence is commonly viewed as a gendered issue that primarily affects women, which tends to leave male victims largely overlooked. This study presents a novel, data-driven analysis of male domestic violence (MDV) in Bangladesh, highlighting the factors that influence it and addressing the challenges posed by a significant categorical imbalance of 5:1 and limited data availability. We collected data from nine major cities in Bangladesh and conducted exploratory data analysis (EDA) to understand the underlying dynamics. EDA revealed patterns such as the high prevalence of verbal abuse, the influence of financial dependency, and the role of familial and socio-economic factors in MDV. To predict and analyze MDV, we implemented 10 traditional machine learning (ML) models, three deep learning models, and two ensemble models, including stacking and hybrid approaches. We propose a stacking ensemble model with ANN and CatBoost as base classifiers and Logistic Regression as the meta-model, which demonstrated the best performance, achieving $95\%$ accuracy, a $99.29\%$ AUC, and balanced metrics across evaluation criteria. Model-specific feature importance analysis of the base classifiers identified key features influencing their decision-making. Model-agnostic explainable AI techniques, such as SHAP and LIME, provided both local and global insights into the decision-making processes of the proposed model, thereby increasing transparency and interpretability. Statistical validation using paired $t$-tests with 10-fold cross-validation and Bonferroni correction ($α= 0.0036$) confirmed the superior performance of our proposed model over alternatives. Our findings challenge the prevailing notion that domestic abuse primarily affects women, emphasizing the need for tailored interventions and support systems for male victims.

cs.CY↗

arXiv2026

Solving Generalized Grouping Problems in Cellular Manufacturing Systems Using a Network Flow Model

This paper focuses on the generalized grouping problem in the context of cellular manufacturing systems (CMS), where parts may have more than one process route. A process route lists the machines corresponding to each part of the operation. Inspired by the extensive and widespread use of network flow algorithms, this research formulates the process route family formation for generalized grouping as a unit capacity minimum cost network flow model. The objective is to minimize dissimilarity (based on the machines required) among the process routes within a family. The proposed model optimally solves the process route family formation problem without pre-specifying the number of part families to be formed. The process route of family formation is the first stage in a hierarchical procedure. For the second stage (machine cell formation), two procedures, a quadratic assignment programming (QAP) formulation, and a heuristic procedure, are proposed. The QAP simultaneously assigns process route families and machines to a pre-specified number of cells in such a way that total machine utilization is maximized. The heuristic procedure for machine cell formation is hierarchical in nature. Computational results for some test problems show that the QAP and the heuristic procedure yield the same results.

cs.AI↗

arXiv2026

Agentic Security: A Systematization of Tools, Failure Modes, and Design Laws for LLM-Driven Penetration Testing

Agentic security uses large-language-model (LLM) agents to plan, dispatch, and interpret security tools. As these systems move from demonstrations to deployed products, practitioners repeatedly encounter the same operational failures. We systematize these failures through a hands-on evaluation of ten widely used static, dynamic, cloud, orchestration, and AI red-teaming tools for unattended pipelines. We introduce a four-dimensional Integration Friction Index that separates one-time engineering cost from recurring organisational, legal, and maintenance cost. We then derive quantitative regularities that explain recurring failure modes. Modelling an agentic security system as stochastic LLM policies wrapped by a deterministic mediator, we show that long-lived sessions lose resident evidence with phase count, while short-lived sub-agents extend the usable horizon according to the compression ratio between raw evidence and its summary. We show that a two-stage verdict cascade multiplies scorer likelihood ratios, but provides little benefit when scorer errors correlate. We show that treating unevaluable outcomes as attack failures biases downstream measurements toward evasive and severe responses. We formulate planner-versus-worker model routing as a knapsack problem and derive a closed-form execution cap for heavy-tailed tools, eta* = alpha v/c. Finally, we show why scope and budget enforcement cannot be delegated to system prompts: prompts do not constrain what actually executes. Inspectra, our implemented platform, serves as a worked instantiation, with mechanisms labelled shipped, partial, or planned, including those that did not work.

cs.CL↗

arXiv2026

TeachMateGPT: A Multi-Agent Knowledge-Grounded Framework for Pedagogical Assessment Generation from Science Curriculum Materials

Automatically generating textbook-grounded assessment items can reduce science teachers' workload, but existing retrieval-augmented generation (RAG) systems rely on flat retrieval, support only single-question generation, lack safeguards against weak evidence, and are ill-suited to low-resource, board-exam-structured curricula. We address these limitations with TeachMateGPT, a multi-agent system contributing four advances to curriculum-grounded science-assessment authoring. (i) COPE, a hierarchical knowledge base replacing token-window chunking with a multi-resolution index that segments documents along syllabus structure and links them at three granularities via a traversable graph-based lineage, matching evidence to each topic's instructional level. (ii) A staged, fail-closed agent pipeline replacing one-shot retrieve-then-generate: routing gates search, retrieval fuses dense and lexical evidence under a coverage gate that withholds generation on insufficient evidence, and specialist agents draft objective and constructed-response items. (iii) SAVER, a source-attributed verification protocol scoring faithfulness, relevance, and hallucination risk against retrieved evidence, applying stricter grounding checks across each creative question's four sub-parts, paired with teacher-in-the-loop evaluation rather than automatic filtering. (iv) NCTB-SciGen8, a curriculum-grounded dataset of 198 items (143 multiple-choice, 55 creative questions) spanning all 14 chapters of the NCTB Class 8 science textbook, produced by the pipeline and rated by three practicing teachers. TeachMateGPT raises faithfulness (0.68 $\rightarrow$ 0.96) and answer relevancy (0.60 $\rightarrow$ 0.89) over a vanilla RAG baseline.

cs.CL↗