Ankit Singh

Quentin Malartic, Nilabhra Roy Chowdhury, Ruxandra Cojocaru et al.

We introduce Falcon2-11B, a foundation model trained on over five trillion tokens, and its multimodal counterpart, Falcon2-11B-vlm, which is a vision-to-text model. We report our findings during the training of the Falcon2-11B which follows a multi-stage approach where the early stages are distinguished by their context length and a final stage where we use a curated, high-quality dataset. Additionally, we report the effect of doubling the batch size mid-training and how training loss spikes are affected by the learning rate. The downstream performance of the foundation model is evaluated on established benchmarks, including multilingual and code datasets. The foundation model shows strong generalization across all the tasks which makes it suitable for downstream finetuning use cases. For the vision language model, we report the performance on several benchmarks and show that our model achieves a higher average score compared to open-source models of similar size. The model weights and code of both Falcon2-11B and Falcon2-11B-vlm are made available under a permissive license.

Mayug Maniparambil, Raiymbek Akshulakov, Yasser Abdelaziz Dahou Djilali et al.

Recent contrastive multimodal vision-language models like CLIP have demonstrated robust open-world semantic understanding, becoming the standard image backbones for vision-language applications. However, recent findings suggest high semantic similarity between well-trained unimodal encoders, which raises a key question: Is there a plausible way to connect unimodal backbones for vision-language tasks? To this end, we propose a novel framework that aligns vision and language using frozen unimodal encoders. It involves selecting semantically similar encoders in the latent space, curating a concept-rich dataset of image-caption pairs, and training simple MLP projectors. We evaluated our approach on 12 zero-shot classification datasets and 2 image-text retrieval datasets. Our best model, utilizing DINOv2 and All-Roberta-Large text encoder, achieves 76\(\%\) accuracy on ImageNet with a 20-fold reduction in data and 65-fold reduction in compute requirements compared multi-modal alignment where models are trained from scratch. The proposed framework enhances the accessibility of multimodal model development while enabling flexible adaptation across diverse scenarios. Code and curated datasets are available at \texttt{github.com/mayug/freeze-align}.

Brigitta Malagurski Törtei, Yasser Dahou, Ngoc Dung Huynh et al.

Vision-Language Models (VLMs) have achieved remarkable progress across tasks such as visual question answering and image captioning. Yet, the extent to which these models perform visual reasoning as opposed to relying on linguistic priors remains unclear. To address this, we introduce VisRes Bench, a benchmark designed to study visual reasoning in naturalistic settings without contextual language supervision. Analyzing model behavior across three levels of complexity, we uncover clear limitations in perceptual and relational visual reasoning capacities. VisRes isolates distinct reasoning abilities across its levels. Level 1 probes perceptual completion and global image matching under perturbations such as blur, texture changes, occlusion, and rotation; Level 2 tests rule-based inference over a single attribute (e.g., color, count, orientation); and Level 3 targets compositional reasoning that requires integrating multiple visual attributes. Across more than 19,000 controlled task images, we find that state-of-the-art VLMs perform near random under subtle perceptual perturbations, revealing limited abstraction beyond pattern recognition. We conclude by discussing how VisRes provides a unified framework for advancing abstract visual reasoning in multimodal research.

Sofian Chaybouti, Sanath Narayan, Yasser Dahou et al.

Vision foundation models trained via multi-teacher distillation offer a promising path toward unified visual representations, yet the learning dynamics and data efficiency of such approaches remain underexplored. In this paper, we systematically study multi-teacher distillation for vision foundation models and identify key factors that enable training at lower computational cost. We introduce Agglomerative Mixture-of-Experts Vision Foundation Models (AMoE), which distill knowledge from SigLIP2 and DINOv3 simultaneously into a Mixture-of-Experts student. We show that (1) our Asymmetric Relation-Knowledge Distillation loss preserves the geometric properties of each teacher while enabling effective knowledge transfer, (2) token-balanced batching that packs varying-resolution images into sequences with uniform token budgets stabilizes representation learning across resolutions without sacrificing performance, and (3) hierarchical clustering and sampling of training data--typically reserved for self-supervised learning--substantially improves sample efficiency over random sampling for multi-teacher distillation. By combining these findings, we curate OpenLVD200M, a 200M-image corpus that demonstrates superior efficiency for multi-teacher distillation. Instantiated in a Mixture-of-Experts. We release OpenLVD200M and distilled models.

Aviraj Bevli, Sofian Chaybouti, Yasser Dahou et al.

Perception-centric systems are typically implemented with a modular encoder-decoder pipeline: a vision backbone for feature extraction and a separate decoder (or late-fusion module) for task prediction. This raises a central question: is this architectural separation essential or can a single early-fusion stack do both perception and task modeling at scale? We introduce Falcon Perception, a unified dense Transformer that processes image patches and text tokens in a shared parameter space from the first layer, using a hybrid attention pattern (bidirectional among image tokens, causal for prediction tokens) to combine global visual context with autoregressive, variable-length instance generation. To keep dense outputs practical, Falcon Perception retains a lightweight token interface and decodes continuous spatial outputs with specialized heads, enabling parallel high-resolution mask prediction. Our design promotes simplicity: we keep a single scalable backbone and shift complexity toward data and training signals, adding only small heads where outputs are continuous and dense. On SA-Co, Falcon Perception improves mask quality to 68.0 Macro-F$_1$ compared to 62.3 of SAM3. We also introduce PBench, a benchmark targeting compositional prompts (OCR, spatial constraints, relations) and dense long-context regimes, where the model shows better gains. Finally, we extend the same early-fusion recipe to Falcon OCR: a compact 300M-parameter model which attains 80.3% on olmOCR and 88.64 on OmniDocBench.

Simone Rossi, Ankit Singh, Thomas Hannagan

The elusive nature of gradient-based optimization in neural networks is tied to their loss landscape geometry, which is poorly understood. However recent work has brought solid evidence that there is essentially no loss barrier between the local solutions of gradient descent, once accounting for weight-permutations that leave the network's computation unchanged. This raises questions for approximate inference in Bayesian neural networks (BNNs), where we are interested in marginalizing over multiple points in the loss landscape. In this work, we first extend the formalism of marginalized loss barrier and solution interpolation to BNNs, before proposing a matching algorithm to search for linearly connected solutions. This is achieved by aligning the distributions of two independent approximate Bayesian solutions with respect to permutation matrices. We build on the results of Ainsworth et al. (2023), reframing the problem as a combinatorial optimization one, using an approximation to the sum of bilinear assignment problem. We then experiment on a variety of architectures and datasets, finding nearly zero marginalized loss barriers for linearly connected solutions.

Long Phan, Alice Gatti, Ziwen Han et al. · amazon-science, apple-ml

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,500 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai.

Aadarsh Sahoo, Ankit Singh, Rameswar Panda et al.

Supervised deep learning methods are enjoying enormous success in many practical applications of computer vision and have the potential to revolutionize robotics. However, the marked performance degradation to biases and imbalanced data questions the reliability of these methods. In this work we address these questions from the perspective of dataset imbalance resulting out of severe under-representation of annotated training data for certain classes and its effect on both deep classification and generation methods. We introduce a joint dataset repairment strategy by combining a neural network classifier with Generative Adversarial Networks (GAN) that makes up for the deficit of training examples from the under-representated class by producing additional training examples. We show that the combined training helps to improve the robustness of both the classifier and the GAN against severe class imbalance. We show the effectiveness of our proposed approach on three very different datasets with different degrees of imbalance in them. The code is available at https://github.com/AadSah/ImbalanceCycleGAN .

Yasser Dahou, Ngoc Dung Huynh, Phuc H. Le-Khac et al.

We present Saliency Benchmark (SalBench), a novel benchmark designed to assess the capability of Large Vision-Language Models (LVLM) in detecting visually salient features that are readily apparent to humans, such as a large circle amidst a grid of smaller ones. This benchmark focuses on low-level features including color, intensity, and orientation, which are fundamental to human visual processing. Our SalBench consists of images that highlight rare, unusual, or unexpected elements within scenes, and naturally draw human attention. It comprises three novel tasks for evaluating the perceptual capabilities of LVLM: Odd-One-Out Detection, Referring Odd-One-Out, and Visual Referring Odd-One-Out. We perform a comprehensive evaluation of state-of-the-art LVLM using SalBench and our findings reveal a surprising limitation: LVLM struggle to identify seemingly obvious visual anomalies, with even the advanced GPT-4o achieving only 47.6\% accuracy on such a simple task. SalBench will be an important step in measuring the capabilities of LVLM that align with the subtle definition of human attention.

Ankit Singh, Efstratios Gavves, Cees G. M. Snoek et al.

Semi-Supervised Learning (SSL) has shown tremendous potential to improve the predictive performance of deep learning models when annotations are hard to obtain. However, the application of SSL has so far been mainly studied in the context of image classification. In this work, we present a semi-supervised approach for spatial-temporal action localization. We introduce a dual guidance network to select better pseudo-bounding boxes. It combines a frame-level classification with a bounding-box prediction to enforce action class consistency across frames and boxes. Our evaluation across well-known spatial-temporal action localization datasets, namely UCF101-24 , J-HMDB-21 and AVA shows that the proposed module considerably enhances the model's performance in limited labeled data settings. Our framework achieves superior results compared to extended image-based semi-supervised baselines.

Ankit Singh

Unsupervised Domain Adaptation (UDA) aims to align the labeled source distribution with the unlabeled target distribution to obtain domain invariant predictive models. However, the application of well-known UDA approaches does not generalize well in Semi-Supervised Domain Adaptation (SSDA) scenarios where few labeled samples from the target domain are available. In this paper, we propose a simple Contrastive Learning framework for semi-supervised Domain Adaptation (CLDA) that attempts to bridge the intra-domain gap between the labeled and unlabeled target distributions and inter-domain gap between source and unlabeled target distribution in SSDA. We suggest employing class-wise contrastive learning to reduce the inter-domain gap and instance-level contrastive alignment between the original (input image) and strongly augmented unlabeled target images to minimize the intra-domain discrepancy. We have shown empirically that both of these modules complement each other to achieve superior performance. Experiments on three well-known domain adaptation benchmark datasets namely DomainNet, Office-Home, and Office31 demonstrate the effectiveness of our approach. CLDA achieves state-of-the-art results on all the above datasets.

Yalda Ghasemi, Ankit Singh, Myunghee Kim et al.

An augmented reality (AR) environment includes a set of digital elements with which the users interact while performing certain tasks. Recent AR head-mounted displays allow users to select how these elements are presented. However, few studies have been conducted to examine the effect of the way of presenting augmented content on user performance and workload. This study aims to evaluate two methods of presenting augmented content - world-locked and head-locked modes in a data entry task. A total of eighteen participants performed the data entry task in this study. The effectiveness of each mode is evaluated in terms of task performance, muscle activity, perceived workload, and usability. The results show that the task completion time is shorter and the typing speed is significantly faster in the head-locked mode while the world-locked mode achieved higher scores in terms of preference. The findings of this study can be applied to AR user interfaces to improve content presentation and enhance the user experience.

Ankit Singh, Omprakash Chakraborty, Ashutosh Varshney et al.

Learning to recognize actions from only a handful of labeled videos is a challenging problem due to the scarcity of tediously collected activity labels. We approach this problem by learning a two-pathway temporal contrastive model using unlabeled videos at two different speeds leveraging the fact that changing video speed does not change an action. Specifically, we propose to maximize the similarity between encoded representations of the same video at two different speeds as well as minimize the similarity between different videos played at different speeds. This way we use the rich supervisory information in terms of `time' that is present in otherwise unsupervised pool of videos. With this simple yet effective strategy of manipulating video playback rates, we considerably outperform video extensions of sophisticated state-of-the-art semi-supervised image recognition methods across multiple diverse benchmark datasets and network architectures. Interestingly, our proposed approach benefits from out-of-domain unlabeled videos showing generalization and robustness. We also perform rigorous ablations and analysis to validate our approach. Project page: https://cvir.github.io/TCL/.

Anirban Santara, Kaustubh Mani, Pranoot Hatwar et al.

Deep learning based landcover classification algorithms have recently been proposed in literature. In hyperspectral images (HSI) they face the challenges of large dimensionality, spatial variability of spectral signatures and scarcity of labeled data. In this article we propose an end-to-end deep learning architecture that extracts band specific spectral-spatial features and performs landcover classification. The architecture has fewer independent connection weights and thus requires lesser number of training data. The method is found to outperform the highest reported accuracies on popular hyperspectral image data sets.

Oualid El Halimi, Trith Patel, Zohaib S. Kiyani et al.

Code cloning is an important software engineering aspect. It is a common software reuse principle that consists of duplicating source code within a program or across different systems owned or maintained by the same entity. There are several contradictory claims concerning the impact of cloning on software stability and maintenance effort. Some papers state that cloning is desired since it speeds up the development process and helps stakeholders meet the tight schedule and deliver on time. Other papers argue that code clone leads to code bloat and causes increase software maintenance costs due to copied defects and dead code. In this paper, we are replicating a previous study done on cloning by the original author. We are repeating his work using the same methods and metrics but with different subjects and experimenters. The paper we are addressing evaluates the impact of code cloning on code stability using three different stability-measuring methods. Our team will apply the same stability measurement techniques on a different software system developed in C programming language to determine generalizability, assure that the results are reliable, validate their outcomes, and to inspire new search by combining previous findings from related studies.