arXiv subjects

Shafin Rahman

Publications and source records attributed to Shafin Rahman.

At least 19 recordsLinked to original sources

arXiv2026

The Gate Always Closes: On Injecting Auxiliary Signals into Frozen Vision-Language Models

Auxiliary signal pathways in VLMs are routinely fitted with learnable gates so the optimiser can decide how much of the signal to admit. We study what the optimiser decides, and whether the model then uses the signal, with hyperbolic visual relational graphs as the experimental vehicle. We find that under cross-entropy fine-tuning the optimiser decides on zero: the gate closes in every design we trained, across five injection designs on LLaVA-1.5-7B, 15 of 15 runs on Qwen2-VL-7B spanning four task families, 3 of 3 seeds on InternVL3-8B, and with LoRA on all linear layers. We attribute this suppression phenomenon to two regimes, a dead-gradient regime formalised through the caption-invariance of image-derived signals and measured directly, and a negative-utility regime in which the auxiliary signal actively hurts the loss. A closed gate does not by itself mean an inert pathway, so we ablate each pathway at inference under one protocol with real region proposals: the norm-aligned designs are behaviourally closed (at most 1.5pp change at the learned gate), whereas the un-normalised gated prefix is open and harmful, costing 6.7pp on relational GQA at its learned gate of +0.005 while helping SugarCrepe. Evaluating every configuration under one inference mode shows that an active injected pathway costs up to 15pp on relational GQA and 7-12pp on out-of-distribution VSR, that the geometric training losses give no measurable in-distribution gain once the pathway is dropped, and that a Euclidean containment control ties the hyperbolic one. Embedding-norm alignment remains necessary for prefix injection that preserves generation. We recommend testing signal use by ablation rather than by reading gate values or attention mass. We open-source the implementation, the checkpoints and the result files behind every table at https://github.com/mrfarazi/the-gate-always-closes.

cs.CV↗

arXiv2026

DUGO: Dual-Scale Gaussian Memory for Online Temporal Action Localization in Egocentric Procedural Videos

Online Temporal Action Localization (OnTAL) aims to detect and localize action instances from untrimmed streaming videos in real time. Unlike offline approaches, OnTAL models have no access to future frames and cannot go back and refine earlier predictions, which makes the task considerably harder given the strict real-time constraints and narrow temporal window available. Most existing OnTAL methods lean on either brief anchor-based context or a single-scale memory of past frames, which limits how well they can capture actions unfolding across multiple time scales or requiring long-range reasoning in Procedural Videos. To address these gaps, we present DUGO (Dual-Scale Encoding and Gaussian Memory Online), a new OnTAL framework that jointly models local motion dynamics and long-range temporal dependencies. At its core, DUGO uses a Dual-Scale Temporal Encoder, pairing depth-wise local representations with global self-attention so it can pick up both fine-grained temporal patterns and broader contextual cues. We also introduce a Gaussian Memory Bank, which represents action memory as stochastic, input-conditioned Gaussian prototypes, allowing for richer temporal memory through gated cross-attention applied at both the encoder and anchor-query stages. Extensive experiments on procedural egocentric datasets (EGTEA, EPIC-Kitchens-100) demonstrate that DUGO achieves state-of-the-art performance while also providing substantial improvements on non-egocentric benchmarks such as Muses. The project page is available at: https://aman554-eq.github.io/DUGO

cs.CV↗

arXiv2026

CAT-GS: Balanced Multimodal Learning via Calibrated Gating and Fusion Surgery

End-to-end training of multimodal neural networks often exhibits unstable neural dynamics characterized by three coupled failure modes that degrade learning: (i) modality imbalance, where one branch dominates gradient-based optimization; (ii) unstable gating, where noisy confidence cues induce erratic modality selection; and (iii) fusion interference, where modality-specific gradients conflict at the shared fusion layer. We propose CAT-GS (Calibrated, Adaptive, Thresholded Gating with Fusion Surgery), a neural dynamics-based optimization controller for intelligent computing applications. CAT-GS operates during backpropagation without modifying model architectures, fusion modules, or task losses. Through calibration of teacher-derived reliability via temperature scaling and EMA smoothing, CAT-GS stabilizes neural dynamics using a margin-thresholded policy to switch between warm-up dropout, weak-modality prioritization, and weak-biased blending, stabilizes gradient magnitudes under aggressive gating via capped gradient-budget renormalization, and applies fusion-only PCGrad to reduce destructive cross-modal interference at the primary shared bottleneck. We evaluate CAT-GS on audio--visual multimodal pattern recognition benchmarks (CREMA-D, AV-MNIST, and VGGSound), a tri-modal setting (UR-FUNNY), controlled synthetic data (CG-MNIST), and additional cross-domain benchmarks (AVE and CMU-MOSI). CAT-GS improves or matches fused multimodal accuracy against strong imbalance-aware baselines (including OGM-GE, G$^2$D, and UMT) across settings, and yields smoother gating behavior with fewer conflicting fusion gradients.

cs.LG↗

arXiv2026

AGSA-Net: Abundance-Guided Self-Attention Network for Spectral Unmixing-Aware Hyperspectral Remote Sensing Image Classification

Hyperspectral image (HSI) classification plays a vital role in remote sensing applications, including agriculture, environmental monitoring, and urban analysis. However, its performance remains challenged by high spectral redundancy, noise sensitivity, and the difficulty of jointly modeling local material composition and long-range spectral dependencies. To address this, we propose AGSA-Net, an abundance-guided self-attention network that explicitly integrates spectral unmixing priors into the classification process. AGSA Net first estimates physically meaningful subpixel abundance maps subject to non-negativity and sum-to-one constraints, regularized by hybrid linear-nonlinear reconstruction decoder. The learned abundances are then used to construct an abundance affinity prior that guides a spectral transformer to emphasize class-discriminative interactions, and the resulting transformer features are fused with compact abundance descriptors for final prediction; in contrast to existing approaches that use abundance as auxiliary or concatenated features. Experiments on Indian Pines, Augsburg, and Berlin demonstrate the benefit of incorporating abundance- guided contextual modeling, particularly in heterogeneous urban scenes. The source code and trained models are available at: https://github.com/nnuvi/AGSA-Net

cs.CV↗

arXiv2026

Beyond Global Editing: Per-Instance Disentangled Subspaces for Training-Free Hallucination Mitigation in LVLMs

Recent advances in large vision-language models (LVLMs) have enabled powerful multimodal reasoning by integrating visual encoders with large language models (LLMs). However, their reliability is frequently undermined by hallucinations, where generated text inaccurately describes the visual input. Although fine-tuning can mitigate this problem, it is computationally expensive and requires large, curated datasets, making training-free alternatives attractive. Among these, model editing is more promising than decoding-based approaches: decoding methods adapt outputs per input but introduce computational overhead and instability, whereas model editing modifies internal representations offline, providing a more efficient and stable solution. However, existing model-editing techniques typically rely on a single global subspace to correct hallucinations, treating all test samples identically and failing to capture diverse hallucination modes across inputs. To address this limitation, we propose a training-free hallucination mitigation framework for dynamic, per-instance suppression at test time. Our method first constructs a set of Disentangled Hallucination Subspaces, each isolating a distinct hallucination mode. During inference, the model adaptively calculates weights reflecting each input's relationship to these subspaces, guiding a dynamically combined projection that selectively suppresses the most probable hallucination directions while preserving image-grounded semantics. Extensive experiments across multiple vision-language benchmarks and LVLM families demonstrate consistent improvements, highlighting the robustness, generalizability, and efficiency of our approach.

cs.CV↗

arXiv2026

CARNet Cycle-Conditioned Core Aggregation and Redistribution for Multivariate Time Series Forecasting

Accurately modeling cross-variate dependencies remains a key challenge in multivariate time series forecasting, particularly in the presence of strong periodic patterns. Many existing approaches rely on attention-based mechanisms that incur quadratic complexity and scale poorly with increasing numbers of variates. Recent attention-free aggregation models address this issue through linear-complexity core-based interactions, but they do not explicitly leverage the global periodic structure present in the data. To overcome this limitation, we propose CARNet, a Cycle-Conditioned Core Aggregation and Redistribution framework that integrates global recurrent cycle information into efficient core based interaction modeling via Multihead Core Aggregation. Extensive experiments on multiple real-world multivariate forecasting benchmarks demonstrate that CARNet consistently outperforms strong transformer and non-attention baselines across diverse prediction horizons while preserving linear-complexity modeling of cross-variate dependencies.

cs.LG↗

arXiv2026

A Point Cloud Transformer for Remote Monitoring and Automated Assessment of Physical Rehabilitation Exercises

Rehabilitation exercises are essential in restoring lost physical functions of patients suffering from various diseases (e.g., Parkinson's, back pain). Carrying out these rehabilitation exercises, often prescribed by health experts, is costly, unavailable, and requires expert supervision. The availability of RGBD images and movement/position data of joints along with expert annotation of exercise data has prompted the use of automatic assessment of the quality of rehabilitation exercises, which is cost-effective and can be carried out at home. However, existing approaches do not extract relevant features, lack practical application, require expensive pre-processing, or overlook crucial features. This study proposes a transformer-based framework for point clouds to extract features and assess rehabilitation exercises by analyzing joint positions collected through RGBD data. We adapt and utilize a curve-based point-cloud feature aggregation technique to augment point-cloud information that aids model output. The transformer architecture also uses axial self-attention, recognizing important joints and their roles to assist users in performing the exercise better. The guided system outperforms existing approaches and is also practically relevant due to its small size, fast inference, and generalization on specific joints in similar exercises. We conduct our experiments on three crucial baseline datasets for rehabilitation exercises: Kimore, UI-PRMD, and IRDS.

cs.CV↗

arXiv2026

One Step Closer to Ground Truth: A Multi-Scale Residual-Aware Representation Learning Pipeline for Predicting Time Series Data

Transformer-based models have emerged as leading paradigms in time-series forecasting in recent years, employing self-attention mechanisms to capture long-range dependencies. Despite their success, these single-stage forecasting architectures exhibit persistent systematic residual biases arising from structural discrepancies, unmodeled stochastic components, or inadequate multi-scale temporal representations. This limitation persists when residuals are treated as irreducible noise, precluding adaptive correction of structured error patterns. To address this limitation, we introduce a two-stage, model-agnostic framework that explicitly decouples forecasting and residual learning into distinct stages of representation learning. A base transformer first generates the initial predictions. Subsequently, a dedicated meta-corrector dynamically models structured error patterns across multivariate channels, preserves cross-variable dependencies, and iteratively refines the residual bias of the base transformer. By formalizing this pipeline as a hypothesis space expansion, our framework addresses approximation limitations inherent in single-stage architectures, removes reliance on restrictive assumptions, and enables end-to-end learning of complex error dynamics. Evaluated on eight popular benchmark datasets using established protocols, our approach achieves state-of-the-art performance, with significant improvements in standard metrics (MSE, MAE). The results demonstrate the framework's ability to mitigate systematic biases and enhance robustness to complex temporal dynamics, advancing the practical applicability of transformer-based forecasting models.

cs.LG↗

arXiv2026

HyperVis: Continuous Latent Visual Relational Graphs on the Lorentz Hyperboloid for Compositional Reasoning

Vision-Language Models (VLMs) struggle with compositional reasoning that requires understanding inter-object relationships. A natural remedy is to inject explicit scene graph triplets $\langle s, p, o \rangle$ from an off-the-shelf scene graph generator (SGG), but we show this backfires: discrete text labels collide with the continuous visual modality, degrading GQA accuracy from 60.38\% to 58.86\%. We propose \textbf{HyperVis}, which bypasses the SGG semantic bottleneck entirely. From $N$ class-agnostic region proposals, we compute a dense $O(N^2)$ visual relation tensor via spatially-biased cross-attention, project it onto a Lorentz hyperboloid, and enforce hierarchy through spatial physics, namely IoA-driven entailment cones and exterior-angle repulsion. We discover that HyperVis contributes in two complementary ways: (1) as a \emph{training-time regularizer}, the hyperbolic relational losses shape LoRA representations that improve generative VQA (GQA 61.03\% vs.\ 57.21\% for LoRA fine-tuning without relational losses, recovering and surpassing the baseline); and (2) as an \emph{inference-time relational encoder}, hyperbolic prefix tokens boost discriminative compositional scoring (SugarCrepe 79.94\%, $+$6.25pp over baseline). The learned curvature stabilises at $κ{=}4.0$, an order of magnitude above prior hyperbolic VLMs where $κ$ typically collapses toward zero, indicating that continuous visual features genuinely require the exponential volume of strongly curved space. A controlled Euclidean ablation confirms this decomposition: the relational pipeline regularises LoRA comparably in flat space (GQA 60.81\%), but the compositionality gain is specifically hyperbolic (SugarCrepe $+$4.58pp over Euclidean), with entailment loss ${\sim}6{\times}$ higher in Euclidean training. Codes are available at TBA.

cs.CV↗

arXiv2026

Privacy-Preserving Empathy Detection in Video Interactions

Detecting empathy from video interactions has emerging applications, yet raw videos that could be used for training AI models are rarely available due to privacy and ethical constraints. Public benchmarks are consequently released only as pre-extracted features, creating a privacy-constrained learning regime whose privacy-utility trade-off is poorly characterised. We formalise three levels of privacy for video-based behavioural prediction -- no privacy (raw video), partial privacy (temporal visual features such as facial landmarks, action units and eye gaze) and strong privacy (summary statistics of those features) -- and ask whether strong, subject-generalisable empathy detection is achievable at the strong-privacy level. We propose TFMPathy, instantiated with two recent Tabular Foundation Models (TFMs) (TabPFN v2 and TabICL), under both in-context learning and fine-tuning paradigms. On a public human-robot interaction benchmark, TFMPathy achieves strong utility under strong privacy, outperforming established baselines by a substantial margin. To assess robustness and facilitate fair, safe deployment, we introduce a cross-subject evaluation protocol that was previously lacking in this benchmark. Under this protocol, TFM fine-tuning improves generalisation capacity substantially (accuracy: $0.590 \rightarrow 0.730$; AUC: $0.564 \rightarrow 0.669$). Aggregating temporal features into summary statistics also suppresses subject-specific and demographic cues, aligning TFMPathy with data-minimisation principles. TFMPathy, therefore, offers a practical route to building AI systems that depend on human-centred video when governance, consent or institutional policies restrict the sharing of raw video. Code will be released upon acceptance at https://github.com/hasan-rakibul/TFMPathy.

cs.CV↗

arXiv2025

Beyond Real Weights: Hypercomplex Representations for Stable Quantization

Multimodal language models (MLLMs) require large parameter capacity to align high-dimensional visual features with linguistic representations, making them computationally heavy and difficult to deploy efficiently. We introduce a progressive reparameterization strategy that compresses these models by gradually replacing dense feed-forward network blocks with compact Parameterized Hypercomplex Multiplication (PHM) layers. A residual interpolation schedule, together with lightweight reconstruction and knowledge distillation losses, ensures that the PHM modules inherit the functional behavior of their dense counterparts during training. This transition yields substantial parameter and FLOP reductions while preserving strong multimodal alignment, enabling faster inference without degrading output quality. We evaluate the approach on multiple vision-language models (VLMs). Our method maintains performance comparable to the base models while delivering significant reductions in model size and inference latency. Progressive PHM substitution thus offers an architecture-compatible path toward more efficient multimodal reasoning and complements existing low-bit quantization techniques.

cs.CV↗

arXiv2025

LAET: A Layer-wise Adaptive Ensemble Tuning Framework for Pretrained Language Models

Natural Language Processing (NLP) has transformed the financial industry, enabling advancements in areas such as textual analysis, risk management, and forecasting. Large language models (LLMs) like BloombergGPT and FinMA have set new benchmarks across various financial NLP tasks, including sentiment analysis, stock movement prediction, and credit risk assessment. Furthermore, FinMA-ES, a bilingual financial LLM, has also demonstrated strong performance using the FLARE and FLARE-ES benchmarks. However, the high computational demands of these models limit the accessibility of many organizations. To address this, we propose Layer-wise Adaptive Ensemble Tuning (LAET), a novel strategy that selectively fine-tunes the most effective layers of pre-trained LLMs by analyzing hidden state representations while freezing less critical layers. LAET significantly reduces computational overhead while enhancing task-specific performance. Our approach shows strong results in financial NLP tasks, outperforming existing benchmarks and state-of-the-art LLMs such as GPT-4, even with smaller LLMs ($\sim$3B parameters). This work bridges cutting-edge financial NLP research and real-world deployment with efficient and scalable models for financial applications.

cs.CL↗

arXiv2025

VideoLights: Feature Refinement and Cross-Task Alignment Transformer for Joint Video Highlight Detection and Moment Retrieval

Prevailing joint prediction transformers for Video Highlight Detection and Moment Retrieval (HD/MR) exhibit deficiencies in handling cross-task dynamics, achieving robust video-text alignment, and utilizing effective attention mechanisms, with the potential of Large Language/Vision-Language Models (LLMs/LVLMs) being largely untapped. This paper introduces VideoLights, a novel HD/MR framework addressing these limitations by incorporating: (i) Convolutional Projection and Feature Refinement modules with an alignment loss for enhanced video-text feature congruity; (ii) a Bi-Directional Cross-Modal Fusion network for strongly coupled query-aware representations; (iii) a Uni-directional joint-task feedback mechanism for synergistic task improvement; (iv) hard positive/negative losses for adaptive learning; and (v) the leveraging of LVLMs (e.g., BLIP-2) for superior multimodal feature integration and intelligent pre-training with synthetic data. Comprehensive evaluations on QVHighlights, TVSum, and Charades-STA benchmarks demonstrate that VideoLights significantly surpasses existing baselines, establishing new state-of-the-art performances. Codes and model checkpoints are available at https://github.com/dpaul06/VideoLights .

cs.CV↗

arXiv2025

UPLME: Uncertainty-Aware Probabilistic Language Modelling for Robust Empathy Regression

Noisy self-reported empathy scores challenge supervised learning for empathy regression. While many algorithms have been proposed for learning with noisy labels in textual classification problems, the regression counterpart is relatively under-explored. We propose UPLME, an uncertainty-aware probabilistic language modelling framework to capture label noise in empathy regression tasks. One of the novelties in UPLME is a probabilistic language model that predicts both empathy scores and heteroscedastic uncertainty, and is trained using Bayesian concepts with variational model ensembling. We further introduce two novel loss components: one penalises degenerate Uncertainty Quantification (UQ), and another enforces similarity between the input pairs on which empathy is being predicted. UPLME achieves state-of-the-art performance (Pearson Correlation Coefficient: $0.558\rightarrow0.580$ and $0.629\rightarrow0.634$) in terms of the performance reported in the literature on two public benchmarks with label noise. Through synthetic label noise injection, we demonstrate that UPLME is effective in distinguishing between noisy and clean samples based on the predicted uncertainty. UPLME further outperform (Calibration error: $0.571\rightarrow0.376$) a recent variational model ensembling-based UQ method designed for regression problems. Code is publicly available at https://github.com/hasan-rakibul/UPLME.

cs.CL↗

arXiv2025

Teacher-Guided One-Shot Pruning via Context-Aware Knowledge Distillation

Unstructured pruning remains a powerful strategy for compressing deep neural networks, yet it often demands iterative train-prune-retrain cycles, resulting in significant computational overhead. To address this challenge, we introduce a novel teacher-guided pruning framework that tightly integrates Knowledge Distillation (KD) with importance score estimation. Unlike prior approaches that apply KD as a post-pruning recovery step, our method leverages gradient signals informed by the teacher during importance score calculation to identify and retain parameters most critical for both task performance and knowledge transfer. Our method facilitates a one-shot global pruning strategy that efficiently eliminates redundant weights while preserving essential representations. After pruning, we employ sparsity-aware retraining with and without KD to recover accuracy without reactivating pruned connections. Comprehensive experiments across multiple image classification benchmarks, including CIFAR-10, CIFAR-100, and TinyImageNet, demonstrate that our method consistently achieves high sparsity levels with minimal performance degradation. Notably, our approach outperforms state-of-the-art baselines such as EPG and EPSD at high sparsity levels, while offering a more computationally efficient alternative to iterative pruning schemes like COLT. The proposed framework offers a computation-efficient, performance-preserving solution well suited for deployment in resource-constrained environments.

cs.CV↗

arXiv2025

Dynamic Temperature Scheduler for Knowledge Distillation

Knowledge Distillation (KD) trains a smaller student model using a large, pre-trained teacher model, with temperature as a key hyperparameter controlling the softness of output probabilities. Traditional methods use a fixed temperature throughout training, which is suboptimal. Moreover, architectural differences between teacher and student often result in mismatched logit magnitudes. We demonstrate that students benefit from softer probabilities early in training but require sharper probabilities in later stages. We introduce Dynamic Temperature Scheduler (DTS), which adjusts temperature dynamically based on the cross-entropy loss gap between teacher and student. To our knowledge, this is the first temperature scheduling method that adapts based on the divergence between teacher and student distributions. Our method integrates seamlessly with existing KD frameworks. We validate DTS across multiple KD strategies on vision (CIFAR-100, Tiny-ImageNet) and NLP tasks (GLUE, Dolly, SelfIns, UnNI, S-NI), consistently outperforming static-temperature baselines. Code is available at https://github.com/Sibgat-Ul/DTS.

cs.LG↗

arXiv2025

Label Semantics for Robust Hyperspectral Image Classification

Hyperspectral imaging (HSI) classification is a critical tool with widespread applications across diverse fields such as agriculture, environmental monitoring, medicine, and materials science. Due to the limited availability of high-quality training samples and the high dimensionality of spectral data, HSI classification models are prone to overfitting and often face challenges in balancing accuracy and computational complexity. Furthermore, most of HSI classification models are monomodal, where it solely relies on spectral-spatial data to learn decision boundaries in the high dimensional embedding space. To address this, we propose a general-purpose Semantic Spectral-Spatial Fusion Network (S3FN) that uses contextual, class specific textual descriptions to complement the training of an HSI classification model. Specifically, S3FN leverages LLMs to generate comprehensive textual descriptions for each class label that captures their unique characteristics and spectral behaviors. These descriptions are then embedded into a vector space using a pre-trained text encoder such as BERT or RoBERTa to extract meaningful label semantics which in turn leads to a better feature-label alignment for improved classification performance. To demonstrate the effectiveness of our approach, we evaluate our model on three diverse HSI benchmark datasets - Hyperspectral Wood, HyperspectralBlueberries, and DeepHS-Fruit and report significant performance boost. Our results highlight the synergy between textual semantics and spectral-spatial data, paving the way for further advancements in semantically augmented HSI classification models. Codes are be available in: https://github.com/milab-nsu/S3FN

cs.CV↗

arXiv2025

HadaSmileNet: Hadamard fusion of handcrafted and deep-learning features for enhancing facial emotion recognition of genuine smiles

The distinction between genuine and posed emotions represents a fundamental pattern recognition challenge with significant implications for data mining applications in social sciences, healthcare, and human-computer interaction. While recent multi-task learning frameworks have shown promise in combining deep learning architectures with handcrafted D-Marker features for smile facial emotion recognition, these approaches exhibit computational inefficiencies due to auxiliary task supervision and complex loss balancing requirements. This paper introduces HadaSmileNet, a novel feature fusion framework that directly integrates transformer-based representations with physiologically grounded D-Markers through parameter-free multiplicative interactions. Through systematic evaluation of 15 fusion strategies, we demonstrate that Hadamard multiplicative fusion achieves optimal performance by enabling direct feature interactions while maintaining computational efficiency. The proposed approach establishes new state-of-the-art results for deep learning methods across four benchmark datasets: UvA-NEMO (88.7 percent, +0.8), MMI (99.7 percent), SPOS (98.5 percent, +0.7), and BBC (100 percent, +5.0). Comprehensive computational analysis reveals 26 percent parameter reduction and simplified training compared to multi-task alternatives, while feature visualization demonstrates enhanced discriminative power through direct domain knowledge integration. The framework's efficiency and effectiveness make it particularly suitable for practical deployment in multimedia data mining applications that require real-time affective computing capabilities.

cs.CV↗