arXiv subjects

Anas Zafar

Publications and source records attributed to Anas Zafar.

14 recordsLinked to original sources

arXiv2026

Do Medical Vision Language Models Actually See? A Counterfactual Grounding Framework and Hard-Negative Contrastive Training for Visually-Reliant Medical VLMs

Large vision language models (VLMs) report strong accuracy on medical question-answering, yet it remains unclear whether they reason from visual evidence or exploit textual shortcuts. We introduce a counterfactual evaluation framework that decouples visual and textual contributions by substituting input images with controlled surrogates blank, pixel-shuffled, image-absent, and CLIP-retrieved hard negatives and derive a suite of grounding metrics including the Visual Reliance Score (VRS) and Visual Hallucination Rate (VHR). We further introduce CORAL (COntrastive Retrieval-Augmented Learning), a 7B-parameter LoRA fine-tune of Qwen2.5-VL-7B trained with a Contrastive Grounding Objective (CGO) that penalises answer invariance under hard-negative image swaps. On a paired controlled evaluation across four closed-form medical VQA benchmarks (PathVQA, PMC-VQA, SLAKE, VQA-RAD; n=400 total), CORAL improves macro accuracy by +6.7 pp (P(Delta>0)=0.988) and reduces VHR by 8.0 pp (P<0.001) over the matched Qwen2.5-VL-7B base; neither MedVLThinker RL variant achieves a significant gain on either metric. Cross-domain diagnostics further reveal that image substitution costs only <=6.5 pp on medical benchmarks versus 48-61 pp on general-domain tasks, situating the grounding gap that CGO targets. We discuss evaluation limitations openly including train/eval benchmark overlap and underpowered secondary metrics and release our framework, training code, and model weights to support reproducible grounding audits of medical VLMs.

cs.CV↗

arXiv2026

Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks

Evaluating large language models (LLMs) for medical applications remains challenging due to benchmark saturation, limited data accessibility, and insufficient coverage of relevant tasks. Existing suites have either saturated, heavily depend on restricted datasets, or lack comprehensive model coverage. We introduce Medmarks, a fully open-source evaluation suite with 30 benchmarks spanning question answering, information extraction, medical calculations, and open-ended clinical reasoning. We perform a systematic evaluation of 61 models across 71 configurations using verifiable metrics and LLM-as-a-Judge. Our results show that frontier reasoning models (Gemini 3 Pro Preview, GPT-5.1, & GPT-5.2) achieve the highest performance across both benchmarks, most frontier proprietary models are significantly more token efficient than open-weight alternatives, medically fine-tuned models outperform their generalist counterparts, and that models are susceptible to answer-order bias (particularly smaller models and Grok 4). A subset of our evals (Medmarks-T) can be directly used as reinforcement learning environments to post-train LLMs for medical reasoning. Code is available at https://github.com/MedARC-AI/Medmarks

cs.CL↗

arXiv2026

Towards Responsible Multimodal Medical Reasoning via Context-Aligned Vision-Language Models

Medical vision-language models (VLMs) show strong performance on radiology tasks but often produce fluent yet weakly grounded conclusions due to over-reliance on a dominant modality. We introduce a context-aligned reasoning framework that enforces agreement across heterogeneous clinical evidence before generating diagnostic conclusions. The proposed approach augments a frozen VLM with structured contextual signals derived from radiomic statistics, explainability activations, and vocabulary-grounded semantic cues. Instead of producing free-form responses, the model generates structured outputs containing supporting evidence, uncertainty estimates, limitations, and safety notes. We observe that auxiliary signals alone provide limited benefit; performance gains emerge only when these signals are integrated through contextual verification. Experiments on chest X-ray datasets demonstrate that context alignment improves discriminative performance (AUC 0.918 to 0.925) while maintaining calibrated uncertainty. The framework also substantially reduces hallucinated keywords (1.14 to 0.25) and produces more concise reasoning explanations (19.4 to 15.3 words) without increasing model confidence (0.70 to 0.68). Cross-dataset evaluation on CheXpert further reveals that modality informativeness significantly influences reasoning behavior. These results suggest that enforcing multi-evidence agreement improves both reliability and trustworthiness in medical multimodal reasoning, while preserving the underlying model architecture.

cs.CV↗

arXiv2026

Projection Guided Personalized Federated Learning for Low Dose CT Denoising

Low-dose CT (LDCT) reduces radiation exposure but introduces protocol-dependent noise and artifacts that vary across institutions. While federated learning enables collaborative training without centralizing patient data, existing methods personalize in image space, making it difficult to separate scanner noise from patient anatomy. We propose ProFed (Projection Guided Personalized Federated Learning), a framework that complements the image space approach by performing dual-level personalization in the projection space, where noise originates during CT measurements before reconstruction combines protocol and anatomy effects. ProFed introduces: (i) anatomy-aware and protocol-aware networks that personalize CT reconstruction to patient and scanner-specific features, (ii) multi-constraint projection losses that enforce consistency with CT measurements, and (iii) uncertainty-guided selective aggregation that weights clients by prediction confidence. Extensive experiments on the Mayo Clinic 2016 dataset demonstrate that ProFed achieves 42.56 dB PSNR with CNN backbones and 44.83 dB with Transformers, outperforming 11 federated learning baselines, including the physics-informed SCAN-PhysFed by +1.42 dB.

eess.IV↗

arXiv2026

Beyond Anatomy: Explainable ASD Classification from rs-fMRI via Functional Parcellation and Graph Attention Networks

Anatomical brain parcellations dominate rs-fMRI-based Autism Spectrum Disorder (ASD) classification, yet their rigid boundaries may fail to capture the idiosyncratic connectivity patterns that characterise ASD. We present a graph-based deep learning framework comparing anatomical (AAL, 116 ROIs) and functionally-derived (MSDL, 39 ROIs) parcellation strategies on the ABIDE I dataset. Our FSL preprocessing pipeline handles multi-site heterogeneity across 400 balanced subjects, with site-stratified 70/15/15 splits to prevent data leakage. Gaussian noise augmentation within training folds expands samples from 280 to 1,680. A three phase pipeline progresses from a baseline GCN with AAL (73.3% accuracy, AUC=0.74), to an optimised GCN with MSDL (84.0%, AUC=0.84), to a Graph Attention Network ensemble achieving 95.0% accuracy (AUC=0.98), outperforming all recent GNN-based benchmarks on ABIDE I. The 10.7-point gain from atlas substitution alone demonstrates that functional parcellation is the most impactful modelling decision. Gradient-based saliency and GNNExplainer analyses converge on the Posterior Cingulate Cortex and Precuneus as core Default Mode Network hubs, validating that model decisions reflect ASD neuropathology rather than acquisition artefacts. All code and datasets will be publicly released upon acceptance.

cs.CV↗

arXiv2026

Beyond Accuracy: Evaluating Visual Grounding In Multimodal Medical Reasoning

Recent work shows that text-only reinforcement learning with verifiable rewards (RLVR) can match or outperform image-text RLVR on multimodal medical VQA benchmarks, suggesting current evaluation protocols may fail to measure causal visual dependence. We introduce a counterfactual evaluation framework using real, blank, and shuffled images across four medical VQA benchmarks: PathVQA, PMC-VQA, SLAKE, and VQA-RAD. Beyond accuracy, we measure Visual Reliance Score (VRS), Image Sensitivity (IS), and introduce Hallucinated Visual Reasoning Rate (HVRR) to detect cases where models generate visual claims despite producing image-invariant answers. Our findings reveal that RLVR improves accuracy while degrading visual grounding: text-only RLVR achieves negative VRS on PathVQA (-0.09), performing better with mismatched images, while image-text RLVR reduces image sensitivity to 39.8% overall despite improving accuracy. On VQA-RAD, both variants achieve 63% accuracy through different mechanisms: text-only RLVR retains 81% performance with blank images, while image-text RLVR shows only 29% image sensitivity. Models generate visual claims in 68-74% of responses, yet 38-43% are ungrounded (HVRR). These findings demonstrate that accuracy-only rewards enable shortcut exploitation, and progress requires grounding-aware evaluation protocols and training objectives that explicitly enforce visual dependence.

cs.CV↗

arXiv2026

CARL-CXR: Continual Adapter-Based Routing for Task-Unknown Chest Radiograph Classification

Clinical deployment of chest radiograph classifiers requires models that can be updated as new datasets become available without retraining on previously observed data or degrading validated performance. We study a task-incremental continual learning setting for chest radiograph classification under task-unknown inference, where heterogeneous chest X-ray datasets arrive sequentially and task identity is unavailable at deployment time. We propose CARL-CXR, a continual adapter-based routing framework that maintains a fixed high-capacity backbone while incrementally introducing lightweight task-specific adapters and classifier heads. A latent task selector operates on adapter-conditioned features to dynamically route each input to the most relevant task pathway, leveraging compact task prototypes and feature-level experience replay to preserve task identity across sequential updates without storing raw images. Experiments on MIMIC-CXR and CheXpert two large-scale datasets with distinct patient populations, imaging devices, and annotation pipelines demonstrate that CARL-CXR achieves minimal catastrophic forgetting (0.012 AUROC drop), representing a 6X and 11X reduction over established continual learning baselines LwF and EWC respectively, while maintaining competitive diagnostic performance (AUROC 0.74). Under task unknown deployment, CARL-CXR outperforms joint training by 12.5 points in routing accuracy (75.0% vs. 62.5%): unlike LwF and EWC, which require explicit task identifiers at inference and provide no routing mechanism.

cs.CV↗

arXiv2025

The Next Layer: Augmenting Foundation Models with Structure-Preserving and Attention-Guided Learning for Local Patches to Global Context Awareness in Computational Pathology

Foundation models have recently emerged as powerful feature extractors in computational pathology, yet they typically omit mechanisms for leveraging the global spatial structure of tissues and the local contextual relationships among diagnostically relevant regions - key elements for understanding the tumor microenvironment. Multiple instance learning (MIL) remains an essential next step following foundation model, designing a framework to aggregate patch-level features into slide-level predictions. We present EAGLE-Net, a structure-preserving, attention-guided MIL architecture designed to augment prediction and interpretability. EAGLE-Net integrates multi-scale absolute spatial encoding to capture global tissue architecture, a top-K neighborhood-aware loss to focus attention on local microenvironments, and background suppression loss to minimize false positives. We benchmarked EAGLE-Net on large pan-cancer datasets, including three cancer types for classification (10,260 slides) and seven cancer types for survival prediction (4,172 slides), using three distinct histology foundation backbones (REMEDIES, Uni-V1, Uni2-h). Across tasks, EAGLE-Net achieved up to 3% higher classification accuracy and the top concordance indices in 6 of 7 cancer types, producing smooth, biologically coherent attention maps that aligned with expert annotations and highlighted invasive fronts, necrosis, and immune infiltration. These results position EAGLE-Net as a generalizable, interpretable framework that complements foundation models, enabling improved biomarker discovery, prognostic modeling, and clinical decision support

q-bio.QM↗

arXiv2025

Thinking Beyond Tokens: From Brain-Inspired Intelligence to Cognitive Foundations for Artificial General Intelligence and its Societal Impact

Can machines truly think, reason and act in domains like humans? This enduring question continues to shape the pursuit of Artificial General Intelligence (AGI). Despite the growing capabilities of models such as GPT-4.5, DeepSeek, Claude 3.5 Sonnet, Phi-4, and Grok 3, which exhibit multimodal fluency and partial reasoning, these systems remain fundamentally limited by their reliance on token-level prediction and lack of grounded agency. This paper offers a cross-disciplinary synthesis of AGI development, spanning artificial intelligence, cognitive neuroscience, psychology, generative models, and agent-based systems. We analyze the architectural and cognitive foundations of general intelligence, highlighting the role of modular reasoning, persistent memory, and multi-agent coordination. In particular, we emphasize the rise of Agentic RAG frameworks that combine retrieval, planning, and dynamic tool use to enable more adaptive behavior. We discuss generalization strategies, including information compression, test-time adaptation, and training-free methods, as critical pathways toward flexible, domain-agnostic intelligence. Vision-Language Models (VLMs) are reexamined not just as perception modules but as evolving interfaces for embodied understanding and collaborative task completion. We also argue that true intelligence arises not from scale alone but from the integration of memory and reasoning: an orchestration of modular, interactive, and self-improving components where compression enables adaptive behavior. Drawing on advances in neurosymbolic systems, reinforcement learning, and cognitive scaffolding, we explore how recent architectures begin to bridge the gap between statistical learning and goal-directed cognition. Finally, we identify key scientific, technical, and ethical challenges on the path to AGI.

cs.AI↗

arXiv2025

Evaluating Multimodal Large Language Models on Educational Textbook Question Answering

Multimodal large language models (MLLMs) have shown success in vision-language tasks, but their ability to reason over complex educational materials remains largely untested. This work presents the first evaluation of state-of-the-art MLLMs, including LLaVA-1.5 and LLaMA 3.2-Vision, on the textbook question answering (TQA) task using the CK12-QA dataset. We introduce a multimodal retrieval-augmented generation (RAG) pipeline to simulate real-world learning by providing relevant lesson paragraphs and diagrams as context. Our zero-shot experiments reveal a critical trade-off: while retrieved context improves LLaVA's performance on text-based questions, it significantly degrades the accuracy of the more powerful LLaMA 3.2-Vision on diagram-based tasks, dropping its validation accuracy from 74.07% to 25.93%. We term this statistically significant phenomenon "catastrophic context interference." Furthermore, fine-tuning highlights architectural differences: LLaMA 3.2-Vision's performance improves to 71.16% on the test set, demonstrating its capacity to learn multimodal integration, whereas LLaVA's performance declines, indicating challenges with generalization. Our results underscore the challenges MLLMs face in modality prioritization and context integration, providing a benchmark and pointing to key directions for developing more robust AI-driven educational tools.

cs.CL↗

arXiv2025

Who is Responsible? The Data, Models, Users or Regulations? A Comprehensive Survey on Responsible Generative AI for a Sustainable Future

Generative AI is rapidly moving from research to deployment, elevating the need for responsible development, evaluation, and governance. We conduct a PRISMA guided review of 232 studies (November 2022 - December 2025), spanning large language models, vision language models, diffusion models, and agentic pipelines. We make four contributions: (1) the first survey bridging governance principles, technical evaluation, and domain deployment across all four system types; (2) a ten-criterion rubric (C1-C10) scoring major AI safety benchmarks on risk-surface coverage, paired with a policy crosswalk mapping benchmarks to regulatory requirements; (3) twelve lifecycle KPIs, explainability guidance for foundation models, and a testbed catalogue; and (4) domain-specific analysis across healthcare, finance, education, arts, agriculture, and defense. Three findings emerge: benchmark coverage is dense for bias and toxicity but sparse for privacy, provenance, deepfakes, and system-level failures in agentic settings; evaluations remain largely static and task local, limiting audit portability; and inconsistent documentation complicates cross-release comparison. We outline a research agenda prioritizing adaptive multimodal evaluation, privacy and provenance testing, deepfake risk assessment, calibration reporting, versioned artifacts, and continuous monitoring. This survey offers a structured path to align generative AI evaluation with governance needs for safe and accountable deployment.

cs.CY↗

arXiv2023

Action Segmentation Using 2D Skeleton Heatmaps and Multi-Modality Fusion

This paper presents a 2D skeleton-based action segmentation method with applications in fine-grained human activity recognition. In contrast with state-of-the-art methods which directly take sequences of 3D skeleton coordinates as inputs and apply Graph Convolutional Networks (GCNs) for spatiotemporal feature learning, our main idea is to use sequences of 2D skeleton heatmaps as inputs and employ Temporal Convolutional Networks (TCNs) to extract spatiotemporal features. Despite lacking 3D information, our approach yields comparable/superior performances and better robustness against missing keypoints than previous methods on action segmentation datasets. Moreover, we improve the performances further by using both 2D skeleton heatmaps and RGB videos as inputs. To our best knowledge, this is the first work to utilize 2D skeleton heatmap inputs and the first work to explore 2D skeleton+RGB fusion for action segmentation.

cs.CV↗

arXiv2023

Permutation-Aware Action Segmentation via Unsupervised Frame-to-Segment Alignment

This paper presents an unsupervised transformer-based framework for temporal activity segmentation which leverages not only frame-level cues but also segment-level cues. This is in contrast with previous methods which often rely on frame-level information only. Our approach begins with a frame-level prediction module which estimates framewise action classes via a transformer encoder. The frame-level prediction module is trained in an unsupervised manner via temporal optimal transport. To exploit segment-level information, we utilize a segment-level prediction module and a frame-to-segment alignment module. The former includes a transformer decoder for estimating video transcripts, while the latter matches frame-level features with segment-level features, yielding permutation-aware segmentation results. Moreover, inspired by temporal optimal transport, we introduce simple-yet-effective pseudo labels for unsupervised training of the above modules. Our experiments on four public datasets, i.e., 50 Salads, YouTube Instructions, Breakfast, and Desktop Assembly show that our approach achieves comparable or better performance than previous methods in unsupervised activity segmentation. Our code and dataset are available on our research website: https://retrocausal.ai/research/.

cs.CV↗

arXiv2022

Brain Tumor Synthetic Data Generation with Adaptive StyleGANs

Generative models have been very successful over the years and have received significant attention for synthetic data generation. As deep learning models are getting more and more complex, they require large amounts of data to perform accurately. In medical image analysis, such generative models play a crucial role as the available data is limited due to challenges related to data privacy, lack of data diversity, or uneven data distributions. In this paper, we present a method to generate brain tumor MRI images using generative adversarial networks. We have utilized StyleGAN2 with ADA methodology to generate high-quality brain MRI with tumors while using a significantly smaller amount of training data when compared to the existing approaches. We use three pre-trained models for transfer learning. Results demonstrate that the proposed method can learn the distributions of brain tumors. Furthermore, the model can generate high-quality synthetic brain MRI with a tumor that can limit the small sample size issues. The approach can addresses the limited data availability by generating realistic-looking brain MRI with tumors. The code is available at: ~\url{https://github.com/rizwanqureshi123/Brain-Tumor-Synthetic-Data}.

eess.IV↗