30 citations · 30 across the 10 of their papers we have counts for
14 papers
AnswerMap: Faithful Spatial Interpretability of VLMs from Answer Posteriors
Mohamed Eltahir, Fardows Adam, Duaa M. Tahir +6
When a VLM answers a visual query, current interpretability tools rely on text rationales, which use a mismatched modality, or on internal read-outs, which originate too early to r…
The Model Knows When to Stop: Training-Free Early Stopping for Long-Context Reading
Muath Alyobi, Mohamed Eltahir, Almoayyad Abuljdail +3
Language models often process long inputs sequentially in chunks, but continuing to read after sufficient evidence has been acquired wastes computation. Existing stopping mechanism…
Recipe-Matching, Not Equivalence
Ali Habibullah, Mohammad Alshiekh, Yazan Alshoibi +2
MathNet-Retrieve asks a retriever to find, for a math problem, a document stating the same problem. An LLM under one fixed prompt writes each gold document and its near-miss distra…
Where LLM Graders Succeed and Break: Evidence from Two Computer-Science Exams
Ali Habibullah, Yazan Alshoibi, Mohammad Alshiekh +2
One long-form exam in a large course costs hundreds of grader-hours, and qualified graders are scarce; LLM graders are a tempting alternative. To show its pitfalls we grade a pract…
PILOT: Policy-Informed Learned Optimization for Adaptive Deep Network Training
Sattam Altuuaim, Lama Ayash, Muhammad Mubashar +1
Despite the central role of optimization in deep learning, most optimizers rely on update structures whose functional form is fixed before training begins. This static design can l…
GridProbe: Posterior-Probing for Adaptive Test-Time Compute in Long-Video VLMs
Mohamed Eltahir, Lama Ayash, Ali Habibullah +2
Long-video understanding in VLMs is bottlenecked by a single monolithic forward pass over thousands of frames at quadratic attention cost. A common mitigation is to first select a…
VideoAtlas: Navigating Long-Form Video in Logarithmic Compute
Mohamed Eltahir, Ali Habibullah, Yazan Alshoibi +3
Extending language models to video introduces two challenges: representation, where existing methods rely on lossy approximations, and long-context, where caption- or agent-based p…
Vote-in-Context: Turning VLMs into Zero-Shot Rank Fusers
Mohamed Eltahir, Ali Habibullah, Lama Ayash +2
In the retrieval domain, candidates' fusion from heterogeneous retrievers is a long-standing challenge, particularly for complex, multi-modal data such as videos. While typical fus…
Class-Agnostic Segmentation Loss and Its Application to Salient Object Detection and Segmentation
Angira Sharma, Naeemullah Khan, Muhammad Mubashar +2
In this paper we present a novel loss function, called class-agnostic segmentation (CAS) loss. With CAS loss the class descriptors are learned during training of the network. We do…
DeformRS: Certifying Input Deformations with Randomized Smoothing
Motasem Alfarra, Adel Bibi, Naeemullah Khan +2
Deep neural networks are vulnerable to input deformations in the form of vector fields of pixel displacements and to other parameterized geometric deformations e.g. translations, r…
Shape-Tailored Deep Neural Networks
Naeemullah Khan, Angira Sharma, Ganesh Sundaramoorthi +1
We present Shape-Tailored Deep Neural Networks (ST-DNN). ST-DNN extend convolutional networks (CNN), which aggregate data from fixed shape (square) neighborhoods, to compute descri…
Class-Agnostic Segmentation Loss and Its Application to Salient Object Detection and Segmentation
Angira Sharma, Naeemullah Khan, Ganesh Sundaramoorthi +1
In this paper we present a novel loss function, called class-agnostic segmentation (CAS) loss. With CAS loss the class descriptors are learned during training of the network. We do…
Continual Learning in Low-rank Orthogonal Subspaces
Arslan Chaudhry, Naeemullah Khan, Puneet K. Dokania +1
In continual learning (CL), a learner is faced with a sequence of tasks, arriving one after the other, and the goal is to remember all the tasks once the continual learning experie…
Coarse-to-Fine Segmentation With Shape-Tailored Scale Spaces
Ganesh Sundaramoorthi, Naeemullah Khan, Byung-Woo Hong
We formulate a general energy and method for segmentation that is designed to have preference for segmenting the coarse structure over the fine structure of the data, without smoot…