PRiSM: Prototype Regularization for Few-Shot VLMs

Authors: Ghassen Baklouti, Omprakash Chakraborty, Jose Dolz, Ismail Ben Ayed

Abstract: Training-free few-shot adaptation methods have gained significant attention recently in the context of Vision-language Models (VLMs). Yet, current benchmarks rely on strong assumptions about the statistics of the adaptation data, e.g., class balance. We question these simplifying assumptions and introduce a more realistic benchmark that varies both the levels of class balance and the effective num… ▽ More Training-free few-shot adaptation methods have gained significant attention recently in the context of Vision-language Models (VLMs). Yet, current benchmarks rely on strong assumptions about the statistics of the adaptation data, e.g., class balance. We question these simplifying assumptions and introduce a more realistic benchmark that varies both the levels of class balance and the effective number of classes in few-shot tasks via Dirichlet sampling. Surprisingly, under our setting, we observe substantial drops in the performances of state-of-the-art methods, more so when the number of labeled samples increases. To mitigate this, we introduce PRiSM, a class-prototype regularization that can be deployed as a plug and play module on top of any existing baseline method, significantly improving performances. Our method optimizes a novel multi-term loss, which includes a regularizer maximizing inter-class pairwise distances, along with additional terms promoting support-feature alignment and fidelity to the baseline prototypes. Furthermore, we introduce an effective and computationally efficient block Majorize-Minimize optimizer for our objective. More specifically, we derive a valid blockwise Lipschitz constant (i.e., a bound on the Hessian's spectral norm), which can be computed efficiently via the Gershgorin circle theorem. Extensive experiments show that PRiSM improves several training-free baselines, with large gains when dealing with severe class imbalance and high numbers of classes. △ Less

Submitted 20 July, 2026; originally announced July 2026.

Quantile Adaptive Temperature Scaling for Confidence Calibration

Authors: Omprakash Chakraborty, Leo Fillioux, Ismail Ben Ayed, Jose Dolz

Abstract: Deep neural networks often produce poorly calibrated confidence estimates, overstating their certainty even when predictions are incorrect. Temperature Scaling remains the most widely used posthoc calibration method due to its simplicity and effectiveness, yet its global, uniform rescaling of logits fails to correct the highly heterogeneous structure of miscalibration observed across the confidenc… ▽ More Deep neural networks often produce poorly calibrated confidence estimates, overstating their certainty even when predictions are incorrect. Temperature Scaling remains the most widely used posthoc calibration method due to its simplicity and effectiveness, yet its global, uniform rescaling of logits fails to correct the highly heterogeneous structure of miscalibration observed across the confidence spectrum. In particular, the largest correctness confidence discrepancies arise in different quantile regions depending on the setting, low confidence predictions, where uncertainty matters most, tend to exhibit the largest correctness confidence discrepancies, which standard TS leaves largely unaddressed. We introduce Quantile Adaptive Temperature Scaling (QaTS), a simple and efficient post hoc calibration method that adapts the temperature as a function of a predictions empirical confidence quantile. By mapping confidences into the quantile space, QaTS normalizes the calibration problem, makes the structure of miscalibration explicit and enables a monotone temperature function that adapts across quantiles while leaving well calibrated high confidence predictions largely unchanged. preserving high confidence behavior. This quantile aware formulation aligns naturally with a reparameterized Expected Calibration Error (ECE) objective and yields a sample wise temperature that is robust across a variety of challenging scenarios, such as class imbalance and distributional shifts. Across a broad range of datasets, architectures, evaluation scenarios and diverse tasks, QaTS consistently, and substantially, outperforms state of the art post hoc calibration methods, delivering more reliable and trustworthy confidence estimates without modifying model predictions. △ Less

Submitted 19 June, 2026; originally announced June 2026.

Are Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web Agents

Authors: Sina Hajimiri, Masih Aminbeidokhti, Jose Dolz, Ismail Ben Ayed, Issam H. Laradji, Spandana Gella, Nicolas Gontier

Abstract: Online web agents often augment a base actor with memory, workflow, or skill modules. These modules can improve performance, but they also consume test-time tokens, a cost rarely reported alongside the actor's inference cost. We study online augmentation, where this overhead is paid on every task, and re-evaluate its benefits under a fixed total inference budget. We compare AWM, ASI, and Reasoning… ▽ More Online web agents often augment a base actor with memory, workflow, or skill modules. These modules can improve performance, but they also consume test-time tokens, a cost rarely reported alongside the actor's inference cost. We study online augmentation, where this overhead is paid on every task, and re-evaluate its benefits under a fixed total inference budget. We compare AWM, ASI, and ReasoningBank with a token-matched vanilla baseline that uses the same budget for additional actor steps. Across three WebArena domains and three models, Gemini 3 Flash, GPT-5.4-mini, and Qwen 3.6-27B, the vanilla baseline matches or surpasses all three augmentation methods in aggregate success rate while often using fewer total tokens. We observe a similar trend on WorkArena-L1 with Qwen 3.6-27B, indicating that the effect extends to enterprise knowledge-work tasks. Our results suggest that skills and workflow memory can be useful in specific domains, but their apparent gains often vanish against a budget-matched actor. We further show that run-to-run variance materially affects outcomes and should be reported as a core evaluation criterion for online web agents. △ Less

Submitted 12 June, 2026; originally announced June 2026.

MolRGen: A Training and Evaluation Setting for De Novo Molecular Generation with Reasonning Models

Authors: Philippe Formont, Maxime Darrin, Ismail Ben Ayed, Pablo Piantanida

Abstract: Recent reasoning-based large language models have shown strong performance on tasks with verifiable outcomes, but their use in de novo molecular generation remains limited by the lack of training environments where rewards can be computed without reference molecules. We introduce MolRGen, a benchmark and molecular verifier for training and evaluating reasoning LLMs on de novo molecular generation.… ▽ More Recent reasoning-based large language models have shown strong performance on tasks with verifiable outcomes, but their use in de novo molecular generation remains limited by the lack of training environments where rewards can be computed without reference molecules. We introduce MolRGen, a benchmark and molecular verifier for training and evaluating reasoning LLMs on de novo molecular generation. MolRGen contains approximately 4,500 protein-pocket targets, resulting in 50k multi-objective optimization prompts combining docking scores with molecular properties such as QED, synthetic accessibility, logP, and physicochemical descriptors. Unlike caption-based generation or molecule-editing benchmarks, MolRGen evaluates molecules proposed from scratch by computing rewards at generation time. We benchmark general-purpose and chemistry-specialized open-source LLMs and introduce a diversity-aware top-k metric to measure whether models can generate a diverse set of high-scoring molecules. Finally, we use the verifier to fine-tune a 128B LLM with GRPO, showing improved performance, at the cost of a diversity-exploitation trade-off. MolRGen provides a scalable testbed for studying verifier-based reasoning and reinforcement learning in molecular design. △ Less

Submitted 11 May, 2026; v1 submitted 18 March, 2026; originally announced March 2026.

Information Maximization for Long-Tailed Semi-Supervised Domain Generalization

Authors: Leo Fillioux, Omprakash Chakraborty, Quentin Gopée, Pierre Marza, Paul-Henry Cournède, Stergios Christodoulidis, Maria Vakalopoulou, Ismail Ben Ayed, Jose Dolz

Abstract: Semi-supervised domain generalization (SSDG) has recently emerged as an appealing alternative to tackle domain generalization when labeled data is scarce but unlabeled samples across domains are abundant. In this work, we identify an important limitation that hampers the deployment of state-of-the-art methods on more challenging but practical scenarios. In particular, state-of-the-art SSDG severel… ▽ More Semi-supervised domain generalization (SSDG) has recently emerged as an appealing alternative to tackle domain generalization when labeled data is scarce but unlabeled samples across domains are abundant. In this work, we identify an important limitation that hampers the deployment of state-of-the-art methods on more challenging but practical scenarios. In particular, state-of-the-art SSDG severely suffers in the presence of long-tailed class distributions, an arguably common situation in real-world settings. To alleviate this limitation, we propose IMaX, a simple yet effective objective based on the well-known InfoMax principle adapted to the SSDG scenario, where the Mutual Information (MI) between the learned features and latent labels is maximized, constrained by the supervision from the labeled samples. Our formulation integrates an α-entropic objective, which mitigates the class-balance bias encoded in the standard marginal entropy term of the MI, thereby better handling arbitrary class distributions. IMaX can be seamlessly plugged into recent state-of-the-art SSDG, consistently enhancing their performance, as demonstrated empirically across two different image modalities. △ Less

Submitted 9 March, 2026; originally announced March 2026.

Locality-Attending Vision Transformer

Authors: Sina Hajimiri, Farzad Beizaee, Fereshteh Shakeri, Christian Desrosiers, Ismail Ben Ayed, Jose Dolz

Abstract: Vision transformers have demonstrated remarkable success in classification by leveraging global self-attention to capture long-range dependencies. However, this same mechanism can obscure fine-grained spatial details crucial for tasks such as segmentation. In this work, we seek to enhance segmentation performance of vision transformers after standard image-level classification training. More speci… ▽ More Vision transformers have demonstrated remarkable success in classification by leveraging global self-attention to capture long-range dependencies. However, this same mechanism can obscure fine-grained spatial details crucial for tasks such as segmentation. In this work, we seek to enhance segmentation performance of vision transformers after standard image-level classification training. More specifically, we present a simple yet effective add-on that improves performance on segmentation tasks while retaining vision transformers' image-level recognition capabilities. In our approach, we modulate the self-attention with a learnable Gaussian kernel that biases the attention toward neighboring patches. We further refine the patch representations to learn better embeddings at patch positions. These modifications encourage tokens to focus on local surroundings and ensure meaningful representations at spatial positions, while still preserving the model's ability to incorporate global information. Experiments demonstrate the effectiveness of our modifications, evidenced by substantial segmentation gains on three benchmarks (e.g., over 6% and 4% on ADE20K for ViT Tiny and Base), without changing the training regime or sacrificing classification performance. The code is available at https://github.com/sinahmr/LocAtViT/. △ Less

Submitted 5 March, 2026; originally announced March 2026.

ORION: ORthonormal Text Encoding for Universal VLM AdaptatION

Authors: Omprakash Chakraborty, Jose Dolz, Ismail Ben Ayed

Abstract: Vision language models (VLMs) have demonstrated remarkable generalization across diverse tasks, yet their performance remains constrained by the quality and geometry of the textual prototypes used to represent classes. Standard zero shot classifiers, derived from frozen text encoders and handcrafted prompts, may yield correlated or weakly separated embeddings that limit task specific discriminabil… ▽ More Vision language models (VLMs) have demonstrated remarkable generalization across diverse tasks, yet their performance remains constrained by the quality and geometry of the textual prototypes used to represent classes. Standard zero shot classifiers, derived from frozen text encoders and handcrafted prompts, may yield correlated or weakly separated embeddings that limit task specific discriminability. We introduce ORION, a text encoder fine tuning framework that improves pretrained VLMs using only class names. Our method optimizes, via low rank adaptation, a novel loss integrating two terms, one promoting pairwise orthogonality between the textual representations of the classes of a given task and the other penalizing deviations from the initial class prototypes. Furthermore, we provide a probabilistic interpretation of our orthogonality penalty, connecting it to the general maximum likelihood estimation (MLE) principle via Huygens theorem. We report extensive experiments on 11 benchmarks and three large VLM backbones, showing that the refined textual embeddings yield powerful replacements for the standard CLIP prototypes. Added as plug and play module on top of various state of the art methods, and across different prediction settings (zero shot, few shot and test time adaptation), ORION improves the performance consistently and significantly. △ Less

Submitted 27 March, 2026; v1 submitted 23 February, 2026; originally announced February 2026.

Histopath-C: Towards Realistic Domain Shifts for Histopathology Vision-Language Adaptation

Authors: Mehrdad Noori, Gustavo Adolfo Vargas Hakim, David Osowiechi, Fereshteh Shakeri, Ali Bahri, Moslem Yazdanpanah, Sahar Dastani, Ismail Ben Ayed, Christian Desrosiers

Abstract: Medical Vision-language models (VLMs) have shown remarkable performances in various medical imaging domains such as histo\-pathology by leveraging pre-trained, contrastive models that exploit visual and textual information. However, histopathology images may exhibit severe domain shifts, such as staining, contamination, blurring, and noise, which may severely degrade the VLM's downstream performan… ▽ More Medical Vision-language models (VLMs) have shown remarkable performances in various medical imaging domains such as histo\-pathology by leveraging pre-trained, contrastive models that exploit visual and textual information. However, histopathology images may exhibit severe domain shifts, such as staining, contamination, blurring, and noise, which may severely degrade the VLM's downstream performance. In this work, we introduce Histopath-C, a new benchmark with realistic synthetic corruptions designed to mimic real-world distribution shifts observed in digital histopathology. Our framework dynamically applies corruptions to any available dataset and evaluates Test-Time Adaptation (TTA) mechanisms on the fly. We then propose LATTE, a transductive, low-rank adaptation strategy that exploits multiple text templates, mitigating the sensitivity of histopathology VLMs to diverse text inputs. Our approach outperforms state-of-the-art TTA methods originally designed for natural images across a breadth of histopathology datasets, demonstrating the effectiveness of our proposed design for robust adaptation in histopathology images. Code and data are available at https://github.com/Mehrdad-Noori/Histopath-C. △ Less

Submitted 18 January, 2026; originally announced January 2026.

Class Adaptive Conformal Training

Authors: Badr-Eddine Marani, Julio Silva-Rodriguez, Ismail Ben Ayed, Maria Vakalopoulou, Stergios Christodoulidis, Jose Dolz

Abstract: Deep neural networks have achieved remarkable success across a variety of tasks, yet they often suffer from unreliable probability estimates. As a result, they can be overconfident in their predictions. Conformal Prediction (CP) offers a principled framework for uncertainty quantification, yielding prediction sets with rigorous coverage guarantees. Existing conformal training methods optimize for… ▽ More Deep neural networks have achieved remarkable success across a variety of tasks, yet they often suffer from unreliable probability estimates. As a result, they can be overconfident in their predictions. Conformal Prediction (CP) offers a principled framework for uncertainty quantification, yielding prediction sets with rigorous coverage guarantees. Existing conformal training methods optimize for overall set size, but shaping the prediction sets in a class-conditional manner is not straightforward and typically requires prior knowledge of the data distribution. In this work, we introduce Class Adaptive Conformal Training (CaCT), which formulates conformal training as an augmented Lagrangian optimization problem that adaptively learns to shape prediction sets class-conditionally without making any distributional assumptions. Experiments on multiple benchmark datasets, including standard and long-tailed image recognition as well as text classification, demonstrate that CaCT consistently outperforms prior conformal training methods, producing significantly smaller and more informative prediction sets while maintaining the desired coverage guarantees. △ Less

Submitted 14 January, 2026; originally announced January 2026.

SoC: Semantic Orthogonal Calibration for Test-Time Prompt Tuning

Authors: Leo Fillioux, Omprakash Chakraborty, Ismail Ben Ayed, Paul-Henry Cournède, Stergios Christodoulidis, Maria Vakalopoulou, Jose Dolz

Abstract: With the increasing adoption of vision-language models (VLMs) in critical decision-making systems such as healthcare or autonomous driving, the calibration of their uncertainty estimates becomes paramount. Yet, this dimension has been largely underexplored in the VLM test-time prompt-tuning (TPT) literature, which has predominantly focused on improving their discriminative performance. Recent stat… ▽ More With the increasing adoption of vision-language models (VLMs) in critical decision-making systems such as healthcare or autonomous driving, the calibration of their uncertainty estimates becomes paramount. Yet, this dimension has been largely underexplored in the VLM test-time prompt-tuning (TPT) literature, which has predominantly focused on improving their discriminative performance. Recent state-of-the-art advocates for enforcing full orthogonality over pairs of text prompt embeddings to enhance separability, and therefore calibration. Nevertheless, as we theoretically show in this work, the inherent gradients from fully orthogonal constraints will strongly push semantically related classes away, ultimately making the model overconfident. Based on our findings, we propose Semantic Orthogonal Calibration (SoC), a Huber-based regularizer that enforces smooth prototype separation while preserving semantic proximity, thereby improving calibration compared to prior orthogonality-based approaches. Across a comprehensive empirical validation, we demonstrate that SoC consistently improves calibration performance, while also maintaining competitive discriminative capabilities. △ Less

Submitted 13 January, 2026; originally announced January 2026.

Learning Task-Agnostic Representations through Multi-Teacher Distillation

Authors: Philippe Formont, Maxime Darrin, Banafsheh Karimian, Jackie CK Cheung, Eric Granger, Ismail Ben Ayed, Mohammadhadi Shateri, Pablo Piantanida

Abstract: Casting complex inputs into tractable representations is a critical step across various fields. Diverse embedding models emerge from differences in architectures, loss functions, input modalities and datasets, each capturing unique aspects of the input. Multi-teacher distillation leverages this diversity to enrich representations but often remains tailored to specific tasks. In this paper, we intr… ▽ More Casting complex inputs into tractable representations is a critical step across various fields. Diverse embedding models emerge from differences in architectures, loss functions, input modalities and datasets, each capturing unique aspects of the input. Multi-teacher distillation leverages this diversity to enrich representations but often remains tailored to specific tasks. In this paper, we introduce a task-agnostic framework based on a ``majority vote" objective function. We demonstrate that this function is bounded by the mutual information between student and teachers' embeddings, leading to a task-agnostic distillation loss that eliminates dependence on task-specific labels or prior knowledge. Our evaluations across text, vision models, and molecular modeling show that our method effectively leverages teacher diversity, resulting in representations enabling better performance for a wide range of downstream tasks such as classification, clustering, or regression. Additionally, we train and release state-of-the-art embedding models, enhancing downstream performance in various modalities. △ Less

Submitted 21 October, 2025; originally announced October 2025.

TRUST: Test-Time Refinement using Uncertainty-Guided SSM Traverses

Authors: Sahar Dastani, Ali Bahri, Gustavo Adolfo Vargas Hakim, Moslem Yazdanpanah, Mehrdad Noori, David Osowiechi, Samuel Barbeau, Ismail Ben Ayed, Herve Lombaert, Christian Desrosiers

Abstract: State Space Models (SSMs) have emerged as efficient alternatives to Vision Transformers (ViTs), with VMamba standing out as a pioneering architecture designed for vision tasks. However, their generalization performance degrades significantly under distribution shifts. To address this limitation, we propose TRUST (Test-Time Refinement using Uncertainty-Guided SSM Traverses), a novel test-time adapt… ▽ More State Space Models (SSMs) have emerged as efficient alternatives to Vision Transformers (ViTs), with VMamba standing out as a pioneering architecture designed for vision tasks. However, their generalization performance degrades significantly under distribution shifts. To address this limitation, we propose TRUST (Test-Time Refinement using Uncertainty-Guided SSM Traverses), a novel test-time adaptation (TTA) method that leverages diverse traversal permutations to generate multiple causal perspectives of the input image. Model predictions serve as pseudo-labels to guide updates of the Mamba-specific parameters, and the adapted weights are averaged to integrate the learned information across traversal scans. Altogether, TRUST is the first approach that explicitly leverages the unique architectural properties of SSMs for adaptation. Experiments on seven benchmarks show that TRUST consistently improves robustness and outperforms existing TTA methods. △ Less

Submitted 7 May, 2026; v1 submitted 26 September, 2025; originally announced September 2025.

Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging

Authors: Moslem Yazdanpanah, Ali Bahri, Mehrdad Noori, Sahar Dastani, Gustavo Adolfo Vargas Hakim, David Osowiechi, Ismail Ben Ayed, Christian Desrosiers

Abstract: Test-time adaptation (TTA) is crucial for mitigating performance degradation caused by distribution shifts in 3D point cloud classification. In this work, we introduce Token Purging (PG), a novel backpropagation-free approach that removes tokens highly affected by domain shifts before they reach attention layers. Unlike existing TTA methods, PG operates at the token level, ensuring robust adaptati… ▽ More Test-time adaptation (TTA) is crucial for mitigating performance degradation caused by distribution shifts in 3D point cloud classification. In this work, we introduce Token Purging (PG), a novel backpropagation-free approach that removes tokens highly affected by domain shifts before they reach attention layers. Unlike existing TTA methods, PG operates at the token level, ensuring robust adaptation without iterative updates. We propose two variants: PG-SP, which leverages source statistics, and PG-SF, a fully source-free version relying on CLS-token-driven adaptation. Extensive evaluations on ModelNet40-C, ShapeNet-C, and ScanObjectNN-C demonstrate that PG-SP achieves an average of +10.3\% higher accuracy than state-of-the-art backpropagation-free methods, while PG-SF sets new benchmarks for source-free adaptation. Moreover, PG is 12.4 times faster and 5.5 times more memory efficient than our baseline, making it suitable for real-world deployment. Code is available at \hyperlink{https://github.com/MosyMosy/Purge-Gate}{https://github.com/MosyMosy/Purge-Gate} △ Less

Submitted 11 September, 2025; originally announced September 2025.

Language-Aware Information Maximization for Transductive Few-Shot CLIP

Authors: Ghassen Baklouti, Maxime Zanella, Ismail Ben Ayed

Abstract: Transductive few-shot learning has triggered an abundant literature focusing on vision-only models, but is still at a nascent stage within the recent context of foundational vision-language models (VLMs). Only a few recent methods addressed the problem, pointing to the potential of tranduction in VLMs and to the need for VLM-tailored methods. Building on this momentum, we leverage information-theo… ▽ More Transductive few-shot learning has triggered an abundant literature focusing on vision-only models, but is still at a nascent stage within the recent context of foundational vision-language models (VLMs). Only a few recent methods addressed the problem, pointing to the potential of tranduction in VLMs and to the need for VLM-tailored methods. Building on this momentum, we leverage information-theoretic concepts and recent progress in parameter-efficient fine-tuning (PEFT), developing a highly competitive transductive few-shot CLIP method. Specifically, we introduce a novel Language-aware Information MaximizatiOn (LIMO) loss integrating three complementary terms: (i) the mutual information between the vision inputs and the textual class descriptions; (ii) a Kullback-Leibler (KL) divergence penalizing deviation of the network's probabilistic outputs from the text-driven zero-shot predictions; and (iii) a standard cross-entropy loss based on the labeled shots. Furthermore, we challenge the commonly followed fine-tuning practices in the context of transductive few-shot learning, and explore PEFT strategies, completely overlooked in this context. Surprisingly, we observe substantial boosts in performances, which points to the potential of adapting a subset of the model's parameters in the transductive few-shot setting. We report comprehensive evaluations, which show that LIMO outperforms the very recent transductive few-shot CLIP methods by a large margin and yields significant gains over the best-performing inductive methods. Our code is publicly available at:\[ \href{https://github.com/ghassenbaklouti/LIMO}{\text{here}} \] △ Less

Submitted 29 August, 2025; originally announced September 2025.

REFLECT: Rectified Flows for Efficient Brain Anomaly Correction Transport

Authors: Farzad Beizaee, Sina Hajimiri, Ismail Ben Ayed, Gregory Lodygensky, Christian Desrosiers, Jose Dolz

Abstract: Unsupervised anomaly detection (UAD) in brain imaging is crucial for identifying pathologies without the need for labeled data. However, accurately localizing anomalies remains challenging due to the intricate structure of brain anatomy and the scarcity of abnormal examples. In this work, we introduce REFLECT, a novel framework that leverages rectified flows to establish a direct, linear trajector… ▽ More Unsupervised anomaly detection (UAD) in brain imaging is crucial for identifying pathologies without the need for labeled data. However, accurately localizing anomalies remains challenging due to the intricate structure of brain anatomy and the scarcity of abnormal examples. In this work, we introduce REFLECT, a novel framework that leverages rectified flows to establish a direct, linear trajectory for correcting abnormal MR images toward a normal distribution. By learning a straight, one-step correction transport map, our method efficiently corrects brain anomalies and can precisely localize anomalies by detecting discrepancies between anomalous input and corrected counterpart. In contrast to the diffusion-based UAD models, which require iterative stochastic sampling, rectified flows provide a direct transport map, enabling single-step inference. Extensive experiments on popular UAD brain segmentation benchmarks demonstrate that REFLECT significantly outperforms state-of-the-art unsupervised anomaly detection methods. The code is available at https://github.com/farzad-bz/REFLECT. △ Less

Submitted 4 August, 2025; originally announced August 2025.

Regularized Low-Rank Adaptation for Few-Shot Organ Segmentation

Authors: Ghassen Baklouti, Julio Silva-Rodríguez, Jose Dolz, Houda Bahig, Ismail Ben Ayed

Abstract: Parameter-efficient fine-tuning (PEFT) of pre-trained foundation models is increasingly attracting interest in medical imaging due to its effectiveness and computational efficiency. Among these methods, Low-Rank Adaptation (LoRA) is a notable approach based on the assumption that the adaptation inherently occurs in a low-dimensional subspace. While it has shown good performance, its implementation… ▽ More Parameter-efficient fine-tuning (PEFT) of pre-trained foundation models is increasingly attracting interest in medical imaging due to its effectiveness and computational efficiency. Among these methods, Low-Rank Adaptation (LoRA) is a notable approach based on the assumption that the adaptation inherently occurs in a low-dimensional subspace. While it has shown good performance, its implementation requires a fixed and unalterable rank, which might be challenging to select given the unique complexities and requirements of each medical imaging downstream task. Inspired by advancements in natural image processing, we introduce a novel approach for medical image segmentation that dynamically adjusts the intrinsic rank during adaptation. Viewing the low-rank representation of the trainable weight matrices as a singular value decomposition, we introduce an l_1 sparsity regularizer to the loss function, and tackle it with a proximal optimizer. The regularizer could be viewed as a penalty on the decomposition rank. Hence, its minimization enables to find task-adapted ranks automatically. Our method is evaluated in a realistic few-shot fine-tuning setting, where we compare it first to the standard LoRA and then to several other PEFT methods across two distinguishable tasks: base organs and novel organs. Our extensive experiments demonstrate the significant performance improvements driven by our method, highlighting its efficiency and robustness against suboptimal rank initialization. Our code is publicly available: https://github.com/ghassenbaklouti/ARENA △ Less

Submitted 21 July, 2025; originally announced July 2025.

ViLU: Learning Vision-Language Uncertainties for Failure Prediction

Authors: Marc Lafon, Yannis Karmim, Julio Silva-Rodríguez, Paul Couairon, Clément Rambour, Raphaël Fournier-Sniehotta, Ismail Ben Ayed, Jose Dolz, Nicolas Thome

Abstract: Reliable Uncertainty Quantification (UQ) and failure prediction remain open challenges for Vision-Language Models (VLMs). We introduce ViLU, a new Vision-Language Uncertainty quantification framework that contextualizes uncertainty estimates by leveraging all task-relevant textual representations. ViLU constructs an uncertainty-aware multi-modal representation by integrating the visual embedding,… ▽ More Reliable Uncertainty Quantification (UQ) and failure prediction remain open challenges for Vision-Language Models (VLMs). We introduce ViLU, a new Vision-Language Uncertainty quantification framework that contextualizes uncertainty estimates by leveraging all task-relevant textual representations. ViLU constructs an uncertainty-aware multi-modal representation by integrating the visual embedding, the predicted textual embedding, and an image-conditioned textual representation via cross-attention. Unlike traditional UQ methods based on loss prediction, ViLU trains an uncertainty predictor as a binary classifier to distinguish correct from incorrect predictions using a weighted binary cross-entropy loss, making it loss-agnostic. In particular, our proposed approach is well-suited for post-hoc settings, where only vision and text embeddings are available without direct access to the model itself. Extensive experiments on diverse datasets show the significant gains of our method compared to state-of-the-art failure prediction methods. We apply our method to standard classification datasets, such as ImageNet-1k, as well as large-scale image-caption datasets like CC12M and LAION-400M. Ablation studies highlight the critical role of our architecture and training in achieving effective uncertainty quantification. Our code is publicly available and can be found here: https://github.com/ykrmm/ViLU. △ Less

Submitted 19 September, 2025; v1 submitted 10 July, 2025; originally announced July 2025.

Trustworthy Few-Shot Transfer of Medical VLMs through Split Conformal Prediction

Authors: Julio Silva-Rodríguez, Ismail Ben Ayed, Jose Dolz

Abstract: Medical vision-language models (VLMs) have demonstrated unprecedented transfer capabilities and are being increasingly adopted for data-efficient image classification. Despite its growing popularity, its reliability aspect remains largely unexplored. This work explores the split conformal prediction (SCP) framework to provide trustworthiness guarantees when transferring such models based on a smal… ▽ More Medical vision-language models (VLMs) have demonstrated unprecedented transfer capabilities and are being increasingly adopted for data-efficient image classification. Despite its growing popularity, its reliability aspect remains largely unexplored. This work explores the split conformal prediction (SCP) framework to provide trustworthiness guarantees when transferring such models based on a small labeled calibration set. Despite its potential, the generalist nature of the VLMs' pre-training could negatively affect the properties of the predicted conformal sets for specific tasks. While common practice in transfer learning for discriminative purposes involves an adaptation stage, we observe that deploying such a solution for conformal purposes is suboptimal since adapting the model using the available calibration data breaks the rigid exchangeability assumptions for test data in SCP. To address this issue, we propose transductive split conformal adaptation (SCA-T), a novel pipeline for transfer learning on conformal scenarios, which performs an unsupervised transductive adaptation jointly on calibration and test data. We present comprehensive experiments utilizing medical VLMs across various image modalities, transfer tasks, and non-conformity scores. Our framework offers consistent gains in efficiency and conditional coverage compared to SCP, maintaining the same empirical guarantees. △ Less

Submitted 20 June, 2025; originally announced June 2025.

Few-Shot, Now for Real: Medical VLMs Adaptation without Balanced Sets or Validation

Authors: Julio Silva-Rodríguez, Fereshteh Shakeri, Houda Bahig, Jose Dolz, Ismail Ben Ayed

Abstract: Vision-language models (VLMs) are gaining attention in medical image analysis. These are pre-trained on large, heterogeneous data sources, yielding rich and transferable representations. Notably, the combination of modality-specialized VLMs with few-shot adaptation has provided fruitful results, enabling the efficient deployment of high-performing solutions. However, previous works on this topic m… ▽ More Vision-language models (VLMs) are gaining attention in medical image analysis. These are pre-trained on large, heterogeneous data sources, yielding rich and transferable representations. Notably, the combination of modality-specialized VLMs with few-shot adaptation has provided fruitful results, enabling the efficient deployment of high-performing solutions. However, previous works on this topic make strong assumptions about the distribution of adaptation data, which are unrealistic in the medical domain. First, prior art assumes access to a balanced support set, a condition that breaks the natural imbalance in disease prevalence found in real-world scenarios. Second, these works typically assume the presence of an additional validation set to fix critical hyper-parameters, which is highly data-inefficient. This work challenges these favorable deployment scenarios and introduces a realistic, imbalanced, validation-free adaptation setting. Our extensive benchmark across various modalities and downstream tasks demonstrates that current methods systematically compromise their performance when operating under realistic conditions, occasionally even performing worse than zero-shot inference. Also, we introduce a training-free linear probe that adaptively blends visual and textual supervision. Detailed studies demonstrate that the proposed solver is a strong, efficient baseline, enabling robust adaptation in challenging scenarios. △ Less

Submitted 20 June, 2025; originally announced June 2025.

Full Conformal Adaptation of Medical Vision-Language Models

Authors: Julio Silva-Rodríguez, Leo Fillioux, Paul-Henry Cournède, Maria Vakalopoulou, Stergios Christodoulidis, Ismail Ben Ayed, Jose Dolz

Abstract: Vision-language models (VLMs) pre-trained at large scale have shown unprecedented transferability capabilities and are being progressively integrated into medical image analysis. Although its discriminative potential has been widely explored, its reliability aspect remains overlooked. This work investigates their behavior under the increasingly popular split conformal prediction (SCP) framework, w… ▽ More Vision-language models (VLMs) pre-trained at large scale have shown unprecedented transferability capabilities and are being progressively integrated into medical image analysis. Although its discriminative potential has been widely explored, its reliability aspect remains overlooked. This work investigates their behavior under the increasingly popular split conformal prediction (SCP) framework, which theoretically guarantees a given error level on output sets by leveraging a labeled calibration set. However, the zero-shot performance of VLMs is inherently limited, and common practice involves few-shot transfer learning pipelines, which cannot absorb the rigid exchangeability assumptions of SCP. To alleviate this issue, we propose full conformal adaptation, a novel setting for jointly adapting and conformalizing pre-trained foundation models, which operates transductively over each test data point using a few-shot adaptation set. Moreover, we complement this framework with SS-Text, a novel training-free linear probe solver for VLMs that alleviates the computational cost of such a transductive approach. We provide comprehensive experiments using 3 different modality-specialized medical VLMs and 9 adaptation tasks. Our framework requires exactly the same data as SCP, and provides consistent relative improvements of up to 27% on set efficiency while maintaining the same coverage guarantees. △ Less

Submitted 6 June, 2025; originally announced June 2025.

Vocabulary-free few-shot learning for Vision-Language Models

Authors: Maxime Zanella, Clément Fuchs, Ismail Ben Ayed, Christophe De Vleeschouwer

Abstract: Recent advances in few-shot adaptation for Vision-Language Models (VLMs) have greatly expanded their ability to generalize across tasks using only a few labeled examples. However, existing approaches primarily build upon the strong zero-shot priors of these models by leveraging carefully designed, task-specific prompts. This dependence on predefined class names can restrict their applicability, es… ▽ More Recent advances in few-shot adaptation for Vision-Language Models (VLMs) have greatly expanded their ability to generalize across tasks using only a few labeled examples. However, existing approaches primarily build upon the strong zero-shot priors of these models by leveraging carefully designed, task-specific prompts. This dependence on predefined class names can restrict their applicability, especially in scenarios where exact class names are unavailable or difficult to specify. To address this limitation, we introduce vocabulary-free few-shot learning for VLMs, a setting where target class instances - that is, images - are available but their corresponding names are not. We propose Similarity Mapping (SiM), a simple yet effective baseline that classifies target instances solely based on similarity scores with a set of generic prompts (textual or visual), eliminating the need for carefully handcrafted prompts. Although conceptually straightforward, SiM demonstrates strong performance, operates with high computational efficiency (learning the mapping typically takes less than one second), and provides interpretability by linking target classes to generic prompts. We believe that our approach could serve as an important baseline for future research in vocabulary-free few-shot learning. Code is available at https://github.com/MaxZanella/vocabulary-free-FSL. △ Less

Submitted 4 June, 2025; originally announced June 2025.

Conformal Prediction for Zero-Shot Models

Authors: Julio Silva-Rodríguez, Ismail Ben Ayed, Jose Dolz

Abstract: Vision-language models pre-trained at large scale have shown unprecedented adaptability and generalization to downstream tasks. Although its discriminative potential has been widely explored, its reliability and uncertainty are still overlooked. In this work, we investigate the capabilities of CLIP models under the split conformal prediction paradigm, which provides theoretical guarantees to black… ▽ More Vision-language models pre-trained at large scale have shown unprecedented adaptability and generalization to downstream tasks. Although its discriminative potential has been widely explored, its reliability and uncertainty are still overlooked. In this work, we investigate the capabilities of CLIP models under the split conformal prediction paradigm, which provides theoretical guarantees to black-box models based on a small, labeled calibration set. In contrast to the main body of literature on conformal predictors in vision classifiers, foundation models exhibit a particular characteristic: they are pre-trained on a one-time basis on an inaccessible source domain, different from the transferred task. This domain drift negatively affects the efficiency of the conformal sets and poses additional challenges. To alleviate this issue, we propose Conf-OT, a transfer learning setting that operates transductive over the combined calibration and query sets. Solving an optimal transport problem, the proposed method bridges the domain gap between pre-training and adaptation without requiring additional data splits but still maintaining coverage guarantees. We comprehensively explore this conformal prediction strategy on a broad span of 15 datasets and three non-conformity scores. Conf-OT provides consistent relative improvements of up to 20% on set efficiency while being 15 times faster than popular transductive approaches. △ Less

Submitted 30 May, 2025; originally announced May 2025.

Recovering Fairness Directly from Modularity: a New Way for Fair Community Partitioning

Authors: Yufeng Wang, Yiguang Bai, Tianqing Zhu, Ismail Ben Ayed, Jing Yuan

Abstract: Community partitioning is crucial in network analysis, with modularity optimization being the prevailing technique. However, traditional modularity-based methods often overlook fairness, a critical aspect in real-world applications. To address this, we introduce protected group networks and propose a novel fairness-modularity metric. This metric extends traditional modularity by explicitly incorpo… ▽ More Community partitioning is crucial in network analysis, with modularity optimization being the prevailing technique. However, traditional modularity-based methods often overlook fairness, a critical aspect in real-world applications. To address this, we introduce protected group networks and propose a novel fairness-modularity metric. This metric extends traditional modularity by explicitly incorporating fairness, and we prove that minimizing it yields naturally fair partitions for protected groups while maintaining theoretical soundness. We develop a general optimization framework for fairness partitioning and design the efficient Fair Fast Newman (FairFN) algorithm, enhancing the Fast Newman (FN) method to optimize both modularity and fairness. Experiments show FairFN achieves significantly improved fairness and high-quality partitions compared to state-of-the-art methods, especially on unbalanced datasets. △ Less

Submitted 27 May, 2025; originally announced May 2025.

Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation

Authors: Mehrdad Noori, David Osowiechi, Gustavo Adolfo Vargas Hakim, Ali Bahri, Moslem Yazdanpanah, Sahar Dastani, Farzad Beizaee, Ismail Ben Ayed, Christian Desrosiers

Abstract: Recently, test-time adaptation has attracted wide interest in the context of vision-language models for image classification. However, to the best of our knowledge, the problem is completely overlooked in dense prediction tasks such as Open-Vocabulary Semantic Segmentation (OVSS). In response, we propose a novel TTA method tailored to adapting VLMs for segmentation during test time. Unlike TTA met… ▽ More Recently, test-time adaptation has attracted wide interest in the context of vision-language models for image classification. However, to the best of our knowledge, the problem is completely overlooked in dense prediction tasks such as Open-Vocabulary Semantic Segmentation (OVSS). In response, we propose a novel TTA method tailored to adapting VLMs for segmentation during test time. Unlike TTA methods for image classification, our Multi-Level and Multi-Prompt (MLMP) entropy minimization integrates features from intermediate vision-encoder layers and is performed with different text-prompt templates at both the global CLS token and local pixel-wise levels. Our approach could be used as plug-and-play for any segmentation network, does not require additional training data or labels, and remains effective even with a single test sample. Furthermore, we introduce a comprehensive OVSS TTA benchmark suite, which integrates a rigorous evaluation protocol, nine segmentation datasets, 15 common synthetic corruptions, and additional real and rendered domain shifts, \textbf{with a total of 87 distinct test scenarios}, establishing a standardized and comprehensive testbed for future TTA research in open-vocabulary segmentation. Our experiments on this suite demonstrate that our segmentation-tailored method consistently delivers significant gains over direct adoption of TTA classification baselines. Code and data are available at https://github.com/dosowiechi/MLMP. △ Less

Submitted 9 November, 2025; v1 submitted 27 May, 2025; originally announced May 2025.

SMART-PC: Skeletal Model Adaptation for Robust Test-Time Training in Point Clouds

Authors: Ali Bahri, Moslem Yazdanpanah, Sahar Dastani, Mehrdad Noori, Gustavo Adolfo Vargas Hakim, David Osowiechi, Farzad Beizaee, Ismail Ben Ayed, Christian Desrosiers

Abstract: Test-Time Training (TTT) has emerged as a promising solution to address distribution shifts in 3D point cloud classification. However, existing methods often rely on computationally expensive backpropagation during adaptation, limiting their applicability in real-world, time-sensitive scenarios. In this paper, we introduce SMART-PC, a skeleton-based framework that enhances resilience to corruption… ▽ More Test-Time Training (TTT) has emerged as a promising solution to address distribution shifts in 3D point cloud classification. However, existing methods often rely on computationally expensive backpropagation during adaptation, limiting their applicability in real-world, time-sensitive scenarios. In this paper, we introduce SMART-PC, a skeleton-based framework that enhances resilience to corruptions by leveraging the geometric structure of 3D point clouds. During pre-training, our method predicts skeletal representations, enabling the model to extract robust and meaningful geometric features that are less sensitive to corruptions, thereby improving adaptability to test-time distribution shifts. Unlike prior approaches, SMART-PC achieves real-time adaptation by eliminating backpropagation and updating only BatchNorm statistics, resulting in a lightweight and efficient framework capable of achieving high frame-per-second rates while maintaining superior classification performance. Extensive experiments on benchmark datasets, including ModelNet40-C, ShapeNet-C, and ScanObjectNN-C, demonstrate that SMART-PC achieves state-of-the-art results, outperforming existing methods such as MATE in terms of both accuracy and computational efficiency. The implementation is available at: https://github.com/AliBahri94/SMART-PC. △ Less

Submitted 26 May, 2025; originally announced May 2025.

Sparsity Outperforms Low-Rank Projections in Few-Shot Adaptation

Authors: Nairouz Mrabah, Nicolas Richet, Ismail Ben Ayed, Éric Granger

Abstract: Adapting Vision-Language Models (VLMs) to new domains with few labeled samples remains a significant challenge due to severe overfitting and computational constraints. State-of-the-art solutions, such as low-rank reparameterization, mitigate these issues but often struggle with generalization and require extensive hyperparameter tuning. In this paper, a novel Sparse Optimization (SO) framework is… ▽ More Adapting Vision-Language Models (VLMs) to new domains with few labeled samples remains a significant challenge due to severe overfitting and computational constraints. State-of-the-art solutions, such as low-rank reparameterization, mitigate these issues but often struggle with generalization and require extensive hyperparameter tuning. In this paper, a novel Sparse Optimization (SO) framework is proposed. Unlike low-rank approaches that typically constrain updates to a fixed subspace, our SO method leverages high sparsity to dynamically adjust very few parameters. We introduce two key paradigms. First, we advocate for \textit{local sparsity and global density}, which updates a minimal subset of parameters per iteration while maintaining overall model expressiveness. As a second paradigm, we advocate for \textit{local randomness and global importance}, which sparsifies the gradient using random selection while pruning the first moment based on importance. This combination significantly mitigates overfitting and ensures stable adaptation in low-data regimes. Extensive experiments on 11 diverse datasets show that SO achieves state-of-the-art few-shot adaptation performance while reducing memory overhead. △ Less

Submitted 11 August, 2025; v1 submitted 16 April, 2025; originally announced April 2025.

Analyzing the Impact of Low-Rank Adaptation for Cross-Domain Few-Shot Object Detection in Aerial Images

Authors: Hicham Talaoubrid, Anissa Mokraoui, Ismail Ben Ayed, Axel Prouvost, Sonimith Hang, Monit Korn, Rémi Harvey

Abstract: This paper investigates the application of Low-Rank Adaptation (LoRA) to small models for cross-domain few-shot object detection in aerial images. Originally designed for large-scale models, LoRA helps mitigate overfitting, making it a promising approach for resource-constrained settings. We integrate LoRA into DiffusionDet, and evaluate its performance on the DOTA and DIOR datasets. Our results s… ▽ More This paper investigates the application of Low-Rank Adaptation (LoRA) to small models for cross-domain few-shot object detection in aerial images. Originally designed for large-scale models, LoRA helps mitigate overfitting, making it a promising approach for resource-constrained settings. We integrate LoRA into DiffusionDet, and evaluate its performance on the DOTA and DIOR datasets. Our results show that LoRA applied after an initial fine-tuning slightly improves performance in low-shot settings (e.g., 1-shot and 5-shot), while full fine-tuning remains more effective in higher-shot configurations. These findings highlight LoRA's potential for efficient adaptation in aerial object detection, encouraging further research into parameter-efficient fine-tuning strategies for few-shot learning. Our code is available here: https://github.com/HichTala/LoRA-DiffusionDet. △ Less

Submitted 8 April, 2025; originally announced April 2025.

A Reality Check of Vision-Language Pre-training in Radiology: Have We Progressed Using Text?

Authors: Julio Silva-Rodríguez, Jose Dolz, Ismail Ben Ayed

Abstract: Vision-language pre-training has recently gained popularity as it allows learning rich feature representations using large-scale data sources. This paradigm has quickly made its way into the medical image analysis community. In particular, there is an impressive amount of recent literature developing vision-language models for radiology. However, the available medical datasets with image-text supe… ▽ More Vision-language pre-training has recently gained popularity as it allows learning rich feature representations using large-scale data sources. This paradigm has quickly made its way into the medical image analysis community. In particular, there is an impressive amount of recent literature developing vision-language models for radiology. However, the available medical datasets with image-text supervision are scarce, and medical concepts are fine-grained, involving expert knowledge that existing vision-language models struggle to encode. In this paper, we propose to take a prudent step back from the literature and revisit supervised, unimodal pre-training, using fine-grained labels instead. We conduct an extensive comparison demonstrating that unimodal pre-training is highly competitive and better suited to integrating heterogeneous data sources. Our results also question the potential of recent vision-language models for open-vocabulary generalization, which have been evaluated using optimistic experimental settings. Finally, we study novel alternatives to better integrate fine-grained labels and noisy text supervision. △ Less

Submitted 7 April, 2025; originally announced April 2025.

Spectral Informed Mamba for Robust Point Cloud Processing

Authors: Ali Bahri, Moslem Yazdanpanah, Mehrdad Noori, Sahar Dastani, Milad Cheraghalikhani, David Osowiechi, Gustavo Adolfo Vargas Hakim, Farzad Beizaee, Ismail Ben Ayed, Christian Desrosiers

Abstract: State space models have shown significant promise in Natural Language Processing (NLP) and, more recently, computer vision. This paper introduces a new methodology leveraging Mamba and Masked Autoencoder networks for point cloud data in both supervised and self-supervised learning. We propose three key contributions to enhance Mamba's capability in processing complex point cloud structures. First,… ▽ More State space models have shown significant promise in Natural Language Processing (NLP) and, more recently, computer vision. This paper introduces a new methodology leveraging Mamba and Masked Autoencoder networks for point cloud data in both supervised and self-supervised learning. We propose three key contributions to enhance Mamba's capability in processing complex point cloud structures. First, we exploit the spectrum of a graph Laplacian to capture patch connectivity, defining an isometry-invariant traversal order that is robust to viewpoints and better captures shape manifolds than traditional 3D grid-based traversals. Second, we adapt segmentation via a recursive patch partitioning strategy informed by Laplacian spectral components, allowing finer integration and segment analysis. Third, we address token placement in Masked Autoencoder for Mamba by restoring tokens to their original positions, which preserves essential order and improves learning. Extensive experiments demonstrate the improvements of our approach in classification, segmentation, and few-shot tasks over state-of-the-art baselines. △ Less

Submitted 25 March, 2025; v1 submitted 6 March, 2025; originally announced March 2025.

Variable Bregman Majorization-Minimization Algorithm and its Application to Dirichlet Maximum Likelihood Estimation

Authors: Ségolène Martin, Jean-Christophe Pesquet, Gabriele Steidl, Ismail Ben Ayed

Abstract: We propose a novel Bregman descent algorithm for minimizing a convex function that is expressed as the sum of a differentiable part (defined over an open set) and a possibly nonsmooth term. The approach, referred to as the Variable Bregman Majorization-Minimization (VBMM) algorithm, extends the Bregman Proximal Gradient method by allowing the Bregman function used in the divergence to adaptively v… ▽ More We propose a novel Bregman descent algorithm for minimizing a convex function that is expressed as the sum of a differentiable part (defined over an open set) and a possibly nonsmooth term. The approach, referred to as the Variable Bregman Majorization-Minimization (VBMM) algorithm, extends the Bregman Proximal Gradient method by allowing the Bregman function used in the divergence to adaptively vary at each iteration, provided it satisfies a majorizing condition on the objective function. This adaptive framework enables the algorithm to approximate the objective more precisely at each iteration, thereby allowing for accelerated convergence compared to the traditional Bregman Proximal Gradient descent. We establish the convergence of the VBMM algorithm to a minimizer under mild assumptions on the family of metrics used. Furthermore, we introduce a novel application of both the Bregman Proximal Gradient method and the VBMM algorithm to the estimation of the multidimensional parameters of a Dirichlet distribution through the maximization of its log-likelihood. Numerical experiments confirm that the VBMM algorithm outperforms existing approaches in terms of convergence speed. △ Less

Submitted 5 February, 2025; v1 submitted 13 January, 2025; originally announced January 2025.

Realistic Test-Time Adaptation of Vision-Language Models

Authors: Maxime Zanella, Clément Fuchs, Christophe De Vleeschouwer, Ismail Ben Ayed

Abstract: The zero-shot capabilities of Vision-Language Models (VLMs) have been widely leveraged to improve predictive performance. However, previous works on transductive or test-time adaptation (TTA) often make strong assumptions about the data distribution, such as the presence of all classes. Our work challenges these favorable deployment scenarios, and introduces a more realistic evaluation framework,… ▽ More The zero-shot capabilities of Vision-Language Models (VLMs) have been widely leveraged to improve predictive performance. However, previous works on transductive or test-time adaptation (TTA) often make strong assumptions about the data distribution, such as the presence of all classes. Our work challenges these favorable deployment scenarios, and introduces a more realistic evaluation framework, including: (i) a variable number of effective classes for adaptation within a single batch, and (ii) non-i.i.d. batches of test samples in online adaptation settings. We provide comprehensive evaluations, comparisons, and ablation studies that demonstrate how current transductive or TTA methods for VLMs systematically compromise the models' initial zero-shot robustness across various realistic scenarios, favoring performance gains under advantageous assumptions about the test samples' distributions. Furthermore, we introduce StatA, a versatile method that could handle a wide range of deployment scenarios, including those with a variable number of effective classes at test time. Our approach incorporates a novel regularization term designed specifically for VLMs, which acts as a statistical anchor preserving the initial text-encoder knowledge, particularly in low-data regimes. Code available at https://github.com/MaxZanella/StatA. △ Less

Submitted 7 January, 2025; originally announced January 2025.

UNEM: UNrolled Generalized EM for Transductive Few-Shot Learning

Authors: Long Zhou, Fereshteh Shakeri, Aymen Sadraoui, Mounir Kaaniche, Jean-Christophe Pesquet, Ismail Ben Ayed

Abstract: Transductive few-shot learning has recently triggered wide attention in computer vision. Yet, current methods introduce key hyper-parameters, which control the prediction statistics of the test batches, such as the level of class balance, affecting performances significantly. Such hyper-parameters are empirically grid-searched over validation data, and their configurations may vary substantially w… ▽ More Transductive few-shot learning has recently triggered wide attention in computer vision. Yet, current methods introduce key hyper-parameters, which control the prediction statistics of the test batches, such as the level of class balance, affecting performances significantly. Such hyper-parameters are empirically grid-searched over validation data, and their configurations may vary substantially with the target dataset and pre-training model, making such empirical searches both sub-optimal and computationally intractable. In this work, we advocate and introduce the unrolling paradigm, also referred to as "learning to optimize", in the context of few-shot learning, thereby learning efficiently and effectively a set of optimized hyper-parameters. Specifically, we unroll a generalization of the ubiquitous Expectation-Maximization (EM) optimizer into a neural network architecture, mapping each of its iterates to a layer and learning a set of key hyper-parameters over validation data. Our unrolling approach covers various statistical feature distributions and pre-training paradigms, including recent foundational vision-language models and standard vision-only classifiers. We report comprehensive experiments, which cover a breadth of fine-grained downstream image classification tasks, showing significant gains brought by the proposed unrolled EM algorithm over iterative variants. The achieved improvements reach up to 10% and 7.5% on vision-only and vision-language benchmarks, respectively. △ Less

Submitted 11 April, 2025; v1 submitted 21 December, 2024; originally announced December 2024.

Are foundation models for computer vision good conformal predictors?

Authors: Leo Fillioux, Julio Silva-Rodríguez, Ismail Ben Ayed, Paul-Henry Cournède, Maria Vakalopoulou, Stergios Christodoulidis, Jose Dolz

Abstract: Recent advances in self-supervision and contrastive learning have brought the performance of foundation models to unprecedented levels in a variety of tasks. Fueled by this progress, these models are becoming the prevailing approach for a wide array of real-world vision problems, including risk-sensitive and high-stakes applications. However, ensuring safe deployment in these scenarios requires a… ▽ More Recent advances in self-supervision and contrastive learning have brought the performance of foundation models to unprecedented levels in a variety of tasks. Fueled by this progress, these models are becoming the prevailing approach for a wide array of real-world vision problems, including risk-sensitive and high-stakes applications. However, ensuring safe deployment in these scenarios requires a more comprehensive understanding of their uncertainty modeling capabilities, which has received little attention. In this work, we delve into the behaviour of vision and vision-language foundation models under Conformal Prediction (CP), a statistical framework that provides theoretical guarantees of marginal coverage of the true class. Across extensive experiments including popular vision classification benchmarks, well-known foundation vision models, and three CP methods, our findings reveal that foundation models are well-suited for conformalization procedures, particularly those integrating Vision Transformers. We also show that calibrating the confidence predictions of these models, a popular strategy to improve their uncertainty quantification, actually leads to efficiency degradation of the conformal set on adaptive CP methods. Furthermore, few-shot adaptation of Vision-Language Models (VLMs) to downstream tasks, whose popularity is surging, enhances conformal scores compared to zero-shot predictions. Last, our empirical study exposes APS as particularly promising in the context of vision foundation models, as it does not violate the marginal coverage guarantees across multiple challenging, yet realistic scenarios. △ Less

Submitted 16 February, 2026; v1 submitted 8 December, 2024; originally announced December 2024.

Semantic Anchor Transport: Robust Test-Time Adaptation for Vision-Language Models

Authors: Shambhavi Mishra, Julio Silva-Rodriguez, Ismail Ben Ayed, Marco Pedersoli, Jose Dolz

Abstract: Large pre-trained vision-language models (VLMs), such as CLIP, have shown unprecedented zero-shot performance across a wide range of tasks. Nevertheless, these models may be unreliable under distributional shifts, as their performance is significantly degraded. In this work, we investigate how to efficiently utilize class text information to mitigate distribution drifts encountered by VLMs during… ▽ More Large pre-trained vision-language models (VLMs), such as CLIP, have shown unprecedented zero-shot performance across a wide range of tasks. Nevertheless, these models may be unreliable under distributional shifts, as their performance is significantly degraded. In this work, we investigate how to efficiently utilize class text information to mitigate distribution drifts encountered by VLMs during inference. In particular, we propose generating pseudo-labels for the noisy test-time samples by aligning visual embeddings with reliable, text-based semantic anchors. Specifically, to maintain the regular structure of the dataset properly, we formulate the problem as a batch-wise label assignment, which is efficiently solved using Optimal Transport. Our method, Semantic Anchor Transport (SAT), utilizes such pseudo-labels as supervisory signals for test-time adaptation, yielding a principled cross-modal alignment solution. Moreover, SAT further leverages heterogeneous textual clues, with a multi-template distillation approach that replicates multi-view contrastive learning strategies in unsupervised representation learning without incurring additional computational complexity. Extensive experiments on multiple popular test-time adaptation benchmarks presenting diverse complexity empirically show the superiority of SAT, achieving consistent performance gains over recent state-of-the-art methods, yet being computationally efficient. △ Less

Submitted 2 January, 2026; v1 submitted 25 November, 2024; originally announced November 2024.

Test-Time Adaptation in Point Clouds: Leveraging Sampling Variation with Weight Averaging

Authors: Ali Bahri, Moslem Yazdanpanah, Mehrdad Noori, Sahar Dastani, Milad Cheraghalikhani, David Osowiech, Farzad Beizaee, Gustavo adolfo. vargas-hakim, Ismail Ben Ayed, Christian Desrosiers

Abstract: Test-Time Adaptation (TTA) addresses distribution shifts during testing by adapting a pretrained model without access to source data. In this work, we propose a novel TTA approach for 3D point cloud classification, combining sampling variation with weight averaging. Our method leverages Farthest Point Sampling (FPS) and K-Nearest Neighbors (KNN) to create multiple point cloud representations, adap… ▽ More Test-Time Adaptation (TTA) addresses distribution shifts during testing by adapting a pretrained model without access to source data. In this work, we propose a novel TTA approach for 3D point cloud classification, combining sampling variation with weight averaging. Our method leverages Farthest Point Sampling (FPS) and K-Nearest Neighbors (KNN) to create multiple point cloud representations, adapting the model for each variation using the TENT algorithm. The final model parameters are obtained by averaging the adapted weights, leading to improved robustness against distribution shifts. Extensive experiments on ModelNet40-C, ShapeNet-C, and ScanObjectNN-C datasets, with different backbones (Point-MAE, PointNet, DGCNN), demonstrate that our approach consistently outperforms existing methods while maintaining minimal resource overhead. The proposed method effectively enhances model generalization and stability in challenging real-world conditions. △ Less

Submitted 29 December, 2024; v1 submitted 1 November, 2024; originally announced November 2024.

Few-shot Adaptation of Medical Vision-Language Models

Authors: Fereshteh Shakeri, Yunshi Huang, Julio Silva-Rodríguez, Houda Bahig, An Tang, Jose Dolz, Ismail Ben Ayed

Abstract: Integrating image and text data through multi-modal learning has emerged as a new approach in medical imaging research, following its successful deployment in computer vision. While considerable efforts have been dedicated to establishing medical foundation models and their zero-shot transfer to downstream tasks, the popular few-shot setting remains relatively unexplored. Following on from the cur… ▽ More Integrating image and text data through multi-modal learning has emerged as a new approach in medical imaging research, following its successful deployment in computer vision. While considerable efforts have been dedicated to establishing medical foundation models and their zero-shot transfer to downstream tasks, the popular few-shot setting remains relatively unexplored. Following on from the currently strong emergence of this setting in computer vision, we introduce the first structured benchmark for adapting medical vision-language models (VLMs) in a strict few-shot regime and investigate various adaptation strategies commonly used in the context of natural images. Furthermore, we evaluate a simple generalization of the linear-probe adaptation baseline, which seeks an optimal blending of the visual prototypes and text embeddings via learnable class-wise multipliers. Surprisingly, such a text-informed linear probe yields competitive performances in comparison to convoluted prompt-learning and adapter-based strategies, while running considerably faster and accommodating the black-box setting. Our extensive experiments span three different medical modalities and specialized foundation models, nine downstream tasks, and several state-of-the-art few-shot adaptation methods. We made our benchmark and code publicly available to trigger further developments in this emergent subject: \url{https://github.com/FereshteShakeri/few-shot-MedVLMs}. △ Less

Submitted 5 September, 2024; originally announced September 2024.

Boosting Vision-Language Models for Histopathology Classification: Predict all at once

Authors: Maxime Zanella, Fereshteh Shakeri, Yunshi Huang, Houda Bahig, Ismail Ben Ayed

Abstract: The development of vision-language models (VLMs) for histo-pathology has shown promising new usages and zero-shot performances. However, current approaches, which decompose large slides into smaller patches, focus solely on inductive classification, i.e., prediction for each patch is made independently of the other patches in the target test data. We extend the capability of these large models by… ▽ More The development of vision-language models (VLMs) for histo-pathology has shown promising new usages and zero-shot performances. However, current approaches, which decompose large slides into smaller patches, focus solely on inductive classification, i.e., prediction for each patch is made independently of the other patches in the target test data. We extend the capability of these large models by introducing a transductive approach. By using text-based predictions and affinity relationships among patches, our approach leverages the strong zero-shot capabilities of these new VLMs without any additional labels. Our experiments cover four histopathology datasets and five different VLMs. Operating solely in the embedding space (i.e., in a black-box setting), our approach is highly efficient, processing $10^5$ patches in just a few seconds, and shows significant accuracy improvements over inductive zero-shot classification. Code available at https://github.com/FereshteShakeri/Histo-TransCLIP. △ Less

Submitted 3 September, 2024; originally announced September 2024.

Enhancing Remote Sensing Vision-Language Models for Zero-Shot Scene Classification

Authors: Karim El Khoury, Maxime Zanella, Benoît Gérin, Tiffanie Godelaine, Benoît Macq, Saïd Mahmoudi, Christophe De Vleeschouwer, Ismail Ben Ayed

Abstract: Vision-Language Models for remote sensing have shown promising uses thanks to their extensive pretraining. However, their conventional usage in zero-shot scene classification methods still involves dividing large images into patches and making independent predictions, i.e., inductive inference, thereby limiting their effectiveness by ignoring valuable contextual information. Our approach tackles t… ▽ More Vision-Language Models for remote sensing have shown promising uses thanks to their extensive pretraining. However, their conventional usage in zero-shot scene classification methods still involves dividing large images into patches and making independent predictions, i.e., inductive inference, thereby limiting their effectiveness by ignoring valuable contextual information. Our approach tackles this issue by utilizing initial predictions based on text prompting and patch affinity relationships from the image encoder to enhance zero-shot capabilities through transductive inference, all without the need for supervision and at a minor computational cost. Experiments on 10 remote sensing datasets with state-of-the-art Vision-Language Models demonstrate significant accuracy improvements over inductive zero-shot classification. Our source code is publicly available on Github: https://github.com/elkhouryk/RS-TransCLIP △ Less

Submitted 7 January, 2025; v1 submitted 1 September, 2024; originally announced September 2024.

Robust Calibration of Large Vision-Language Adapters

Authors: Balamurali Murugesan, Julio Silva-Rodriguez, Ismail Ben Ayed, Jose Dolz

Abstract: This paper addresses the critical issue of miscalibration in CLIP-based model adaptation, particularly in the challenging scenario of out-of-distribution (OOD) samples, which has been overlooked in the existing literature on CLIP adaptation. We empirically demonstrate that popular CLIP adaptation approaches, such as Adapters, Prompt Learning, and Test-Time Adaptation, substantially degrade the cal… ▽ More This paper addresses the critical issue of miscalibration in CLIP-based model adaptation, particularly in the challenging scenario of out-of-distribution (OOD) samples, which has been overlooked in the existing literature on CLIP adaptation. We empirically demonstrate that popular CLIP adaptation approaches, such as Adapters, Prompt Learning, and Test-Time Adaptation, substantially degrade the calibration capabilities of the zero-shot baseline in the presence of distributional drift. We identify the increase in logit ranges as the underlying cause of miscalibration of CLIP adaptation methods, contrasting with previous work on calibrating fully-supervised models. Motivated by these observations, we present a simple and model-agnostic solution to mitigate miscalibration, by scaling the logit range of each sample to its zero-shot prediction logits. We explore three different alternatives to achieve this, which can be either integrated during adaptation or directly used at inference time. Comprehensive experiments on popular OOD classification benchmarks demonstrate the effectiveness of the proposed approaches in mitigating miscalibration while maintaining discriminative performance, whose improvements are consistent across the three families of these increasingly popular approaches. The code is publicly available at: https://github.com/Bala93/CLIPCalib △ Less

Submitted 18 July, 2024; originally announced July 2024.

FDS: Feedback-guided Domain Synthesis with Multi-Source Conditional Diffusion Models for Domain Generalization

Authors: Mehrdad Noori, Milad Cheraghalikhani, Ali Bahri, Gustavo Adolfo Vargas Hakim, David Osowiechi, Moslem Yazdanpanah, Ismail Ben Ayed, Christian Desrosiers

Abstract: Domain Generalization techniques aim to enhance model robustness by simulating novel data distributions during training, typically through various augmentation or stylization strategies. However, these methods frequently suffer from limited control over the diversity of generated images and lack assurance that these images span distinct distributions. To address these challenges, we propose FDS, F… ▽ More Domain Generalization techniques aim to enhance model robustness by simulating novel data distributions during training, typically through various augmentation or stylization strategies. However, these methods frequently suffer from limited control over the diversity of generated images and lack assurance that these images span distinct distributions. To address these challenges, we propose FDS, Feedback-guided Domain Synthesis, a novel strategy that employs diffusion models to synthesize novel, pseudo-domains by training a single model on all source domains and performing domain mixing based on learned features. By incorporating images that pose classification challenges to models trained on original samples, alongside the original dataset, we ensure the generation of a training set that spans a broad distribution spectrum. Our comprehensive evaluations demonstrate that this methodology sets new benchmarks in domain generalization performance across a range of challenging datasets, effectively managing diverse types of domain shifts. The code can be found at: \url{https://github.com/Mehrdad-Noori/FDS.git}. △ Less

Submitted 18 December, 2024; v1 submitted 3 July, 2024; originally announced July 2024.

WATT: Weight Average Test-Time Adaptation of CLIP

Authors: David Osowiechi, Mehrdad Noori, Gustavo Adolfo Vargas Hakim, Moslem Yazdanpanah, Ali Bahri, Milad Cheraghalikhani, Sahar Dastani, Farzad Beizaee, Ismail Ben Ayed, Christian Desrosiers

Abstract: Vision-Language Models (VLMs) such as CLIP have yielded unprecedented performance for zero-shot image classification, yet their generalization capability may still be seriously challenged when confronted to domain shifts. In response, we present Weight Average Test-Time Adaptation (WATT) of CLIP, a pioneering approach facilitating full test-time adaptation (TTA) of this VLM. Our method employs a d… ▽ More Vision-Language Models (VLMs) such as CLIP have yielded unprecedented performance for zero-shot image classification, yet their generalization capability may still be seriously challenged when confronted to domain shifts. In response, we present Weight Average Test-Time Adaptation (WATT) of CLIP, a pioneering approach facilitating full test-time adaptation (TTA) of this VLM. Our method employs a diverse set of templates for text prompts, augmenting the existing framework of CLIP. Predictions are utilized as pseudo labels for model updates, followed by weight averaging to consolidate the learned information globally. Furthermore, we introduce a text ensemble strategy, enhancing overall test performance by aggregating diverse textual cues. Our findings underscore the efficacy of WATT in enhancing performance across diverse datasets, including CIFAR-10-C, CIFAR-10.1, CIFAR-100-C, VisDA-C, and several other challenging datasets, effectively covering a wide range of domain shifts. Notably, these enhancements are achieved without necessitating additional model transformations or trainable modules. Moreover, compared to other Test-Time Adaptation methods, our approach can operate effectively with just a single image. Highlighting the potential of innovative test-time strategies, this research emphasizes their role in fortifying the adaptability of VLMs. The implementation is available at: \url{https://github.com/Mehrdad-Noori/WATT.git}. △ Less

Submitted 24 June, 2024; v1 submitted 19 June, 2024; originally announced June 2024.

When is an Embedding Model More Promising than Another?

Authors: Maxime Darrin, Philippe Formont, Ismail Ben Ayed, Jackie CK Cheung, Pablo Piantanida

Abstract: Embedders play a central role in machine learning, projecting any object into numerical representations that can, in turn, be leveraged to perform various downstream tasks. The evaluation of embedding models typically depends on domain-specific empirical approaches utilizing downstream tasks, primarily because of the lack of a standardized framework for comparison. However, acquiring adequately la… ▽ More Embedders play a central role in machine learning, projecting any object into numerical representations that can, in turn, be leveraged to perform various downstream tasks. The evaluation of embedding models typically depends on domain-specific empirical approaches utilizing downstream tasks, primarily because of the lack of a standardized framework for comparison. However, acquiring adequately large and representative datasets for conducting these assessments is not always viable and can prove to be prohibitively expensive and time-consuming. In this paper, we present a unified approach to evaluate embedders. First, we establish theoretical foundations for comparing embedding models, drawing upon the concepts of sufficiency and informativeness. We then leverage these concepts to devise a tractable comparison criterion (information sufficiency), leading to a task-agnostic and self-supervised ranking procedure. We demonstrate experimentally that our approach aligns closely with the capability of embedding models to facilitate various downstream tasks in both natural language processing and molecular biology. This effectively offers practitioners a valuable tool for prioritizing model trials. △ Less

Submitted 16 November, 2024; v1 submitted 11 June, 2024; originally announced June 2024.

Boosting Vision-Language Models with Transduction

Authors: Maxime Zanella, Benoît Gérin, Ismail Ben Ayed

Abstract: Transduction is a powerful paradigm that leverages the structure of unlabeled data to boost predictive accuracy. We present TransCLIP, a novel and computationally efficient transductive approach designed for Vision-Language Models (VLMs). TransCLIP is applicable as a plug-and-play module on top of popular inductive zero- and few-shot models, consistently improving their performances. Our new objec… ▽ More Transduction is a powerful paradigm that leverages the structure of unlabeled data to boost predictive accuracy. We present TransCLIP, a novel and computationally efficient transductive approach designed for Vision-Language Models (VLMs). TransCLIP is applicable as a plug-and-play module on top of popular inductive zero- and few-shot models, consistently improving their performances. Our new objective function can be viewed as a regularized maximum-likelihood estimation, constrained by a KL divergence penalty that integrates the text-encoder knowledge and guides the transductive learning process. We further derive an iterative Block Majorize-Minimize (BMM) procedure for optimizing our objective, with guaranteed convergence and decoupled sample-assignment updates, yielding computationally efficient transduction for large-scale datasets. We report comprehensive evaluations, comparisons, and ablation studies that demonstrate: (i) Transduction can greatly enhance the generalization capabilities of inductive pretrained zero- and few-shot VLMs; (ii) TransCLIP substantially outperforms standard transductive few-shot learning methods relying solely on vision features, notably due to the KL-based language constraint. △ Less

Submitted 3 June, 2024; originally announced June 2024.

Low-Rank Few-Shot Adaptation of Vision-Language Models

Authors: Maxime Zanella, Ismail Ben Ayed

Abstract: Recent progress in the few-shot adaptation of Vision-Language Models (VLMs) has further pushed their generalization capabilities, at the expense of just a few labeled samples within the target downstream task. However, this promising, already quite abundant few-shot literature has focused principally on prompt learning and, to a lesser extent, on adapters, overlooking the recent advances in Parame… ▽ More Recent progress in the few-shot adaptation of Vision-Language Models (VLMs) has further pushed their generalization capabilities, at the expense of just a few labeled samples within the target downstream task. However, this promising, already quite abundant few-shot literature has focused principally on prompt learning and, to a lesser extent, on adapters, overlooking the recent advances in Parameter-Efficient Fine-Tuning (PEFT). Furthermore, existing few-shot learning methods for VLMs often rely on heavy training procedures and/or carefully chosen, task-specific hyper-parameters, which might impede their applicability. In response, we introduce Low-Rank Adaptation (LoRA) in few-shot learning for VLMs, and show its potential on 11 datasets, in comparison to current state-of-the-art prompt- and adapter-based approaches. Surprisingly, our simple CLIP-LoRA method exhibits substantial improvements, while reducing the training times and keeping the same hyper-parameters in all the target tasks, i.e., across all the datasets and numbers of shots. Certainly, our surprising results do not dismiss the potential of prompt-learning and adapter-based research. However, we believe that our strong baseline could be used to evaluate progress in these emergent subjects in few-shot VLMs. △ Less

Submitted 1 June, 2024; v1 submitted 28 May, 2024; originally announced May 2024.

Transductive Zero-Shot and Few-Shot CLIP

Authors: Ségolène Martin, Yunshi Huang, Fereshteh Shakeri, Jean-Christophe Pesquet, Ismail Ben Ayed

Abstract: Transductive inference has been widely investigated in few-shot image classification, but completely overlooked in the recent, fast growing literature on adapting vision-langage models like CLIP. This paper addresses the transductive zero-shot and few-shot CLIP classification challenge, in which inference is performed jointly across a mini-batch of unlabeled query samples, rather than treating eac… ▽ More Transductive inference has been widely investigated in few-shot image classification, but completely overlooked in the recent, fast growing literature on adapting vision-langage models like CLIP. This paper addresses the transductive zero-shot and few-shot CLIP classification challenge, in which inference is performed jointly across a mini-batch of unlabeled query samples, rather than treating each instance independently. We initially construct informative vision-text probability features, leading to a classification problem on the unit simplex set. Inspired by Expectation-Maximization (EM), our optimization-based classification objective models the data probability distribution for each class using a Dirichlet law. The minimization problem is then tackled with a novel block Majorization-Minimization algorithm, which simultaneously estimates the distribution parameters and class assignments. Extensive numerical experiments on 11 datasets underscore the benefits and efficacy of our batch inference approach.On zero-shot tasks with test batches of 75 samples, our approach yields near 20% improvement in ImageNet accuracy over CLIP's zero-shot performance. Additionally, we outperform state-of-the-art methods in the few-shot setting. The code is available at: https://github.com/SegoleneMartin/transductive-CLIP. △ Less

Submitted 8 April, 2024; originally announced May 2024.

GeoMask3D: Geometrically Informed Mask Selection for Self-Supervised Point Cloud Learning in 3D

Authors: Ali Bahri, Moslem Yazdanpanah, Mehrdad Noori, Milad Cheraghalikhani, Gustavo Adolfo Vargas Hakim, David Osowiechi, Farzad Beizaee, Ismail Ben Ayed, Christian Desrosiers

Abstract: We introduce a pioneering approach to self-supervised learning for point clouds, employing a geometrically informed mask selection strategy called GeoMask3D (GM3D) to boost the efficiency of Masked Auto Encoders (MAE). Unlike the conventional method of random masking, our technique utilizes a teacher-student model to focus on intricate areas within the data, guiding the model's focus toward region… ▽ More We introduce a pioneering approach to self-supervised learning for point clouds, employing a geometrically informed mask selection strategy called GeoMask3D (GM3D) to boost the efficiency of Masked Auto Encoders (MAE). Unlike the conventional method of random masking, our technique utilizes a teacher-student model to focus on intricate areas within the data, guiding the model's focus toward regions with higher geometric complexity. This strategy is grounded in the hypothesis that concentrating on harder patches yields a more robust feature representation, as evidenced by the improved performance on downstream tasks. Our method also presents a complete-to-partial feature-level knowledge distillation technique designed to guide the prediction of geometric complexity utilizing a comprehensive context from feature-level information. Extensive experiments confirm our method's superiority over State-Of-The-Art (SOTA) baselines, demonstrating marked improvements in classification, and few-shot tasks. △ Less

Submitted 17 March, 2025; v1 submitted 20 May, 2024; originally announced May 2024.

On the test-time zero-shot generalization of vision-language models: Do we really need prompt learning?

Authors: Maxime Zanella, Ismail Ben Ayed

Abstract: The development of large vision-language models, notably CLIP, has catalyzed research into effective adaptation techniques, with a particular focus on soft prompt tuning. Conjointly, test-time augmentation, which utilizes multiple augmented views of a single image to enhance zero-shot generalization, is emerging as a significant area of interest. This has predominantly directed research efforts to… ▽ More The development of large vision-language models, notably CLIP, has catalyzed research into effective adaptation techniques, with a particular focus on soft prompt tuning. Conjointly, test-time augmentation, which utilizes multiple augmented views of a single image to enhance zero-shot generalization, is emerging as a significant area of interest. This has predominantly directed research efforts toward test-time prompt tuning. In contrast, we introduce a robust MeanShift for Test-time Augmentation (MTA), which surpasses prompt-based methods without requiring this intensive training procedure. This positions MTA as an ideal solution for both standalone and API-based applications. Additionally, our method does not rely on ad hoc rules (e.g., confidence threshold) used in some previous test-time augmentation techniques to filter the augmented views. Instead, MTA incorporates a quality assessment variable for each view directly into its optimization process, termed as the inlierness score. This score is jointly optimized with a density mode seeking process, leading to an efficient training- and hyperparameter-free approach. We extensively benchmark our method on 15 datasets and demonstrate MTA's superiority and computational efficiency. Deployed easily as plug-and-play module on top of zero-shot models and state-of-the-art few-shot methods, MTA shows systematic and consistent improvements. △ Less

Submitted 3 May, 2024; originally announced May 2024.

CLIPArTT: Adaptation of CLIP to New Domains at Test Time

Authors: Gustavo Adolfo Vargas Hakim, David Osowiechi, Mehrdad Noori, Milad Cheraghalikhani, Ali Bahri, Moslem Yazdanpanah, Ismail Ben Ayed, Christian Desrosiers

Abstract: Pre-trained vision-language models (VLMs), exemplified by CLIP, demonstrate remarkable adaptability across zero-shot classification tasks without additional training. However, their performance diminishes in the presence of domain shifts. In this study, we introduce CLIP Adaptation duRing Test-Time (CLIPArTT), a fully test-time adaptation (TTA) approach for CLIP, which involves automatic text prom… ▽ More Pre-trained vision-language models (VLMs), exemplified by CLIP, demonstrate remarkable adaptability across zero-shot classification tasks without additional training. However, their performance diminishes in the presence of domain shifts. In this study, we introduce CLIP Adaptation duRing Test-Time (CLIPArTT), a fully test-time adaptation (TTA) approach for CLIP, which involves automatic text prompts construction during inference for their use as text supervision. Our method employs a unique, minimally invasive text prompt tuning process, wherein multiple predicted classes are aggregated into a single new text prompt, used as \emph{pseudo label} to re-classify inputs in a transductive manner. Additionally, we pioneer the standardization of TTA benchmarks (e.g., TENT) in the realm of VLMs. Our findings demonstrate that, without requiring additional transformations nor new trainable modules, CLIPArTT enhances performance dynamically across non-corrupted datasets such as CIFAR-100, corrupted datasets like CIFAR-100-C and ImageNet-C, alongside synthetic datasets such as VisDA-C. This research underscores the potential for improving VLMs' adaptability through novel test-time strategies, offering insights for robust performance across varied datasets and environments. The code can be found at: https://github.com/dosowiechi/CLIPArTT.git △ Less

Submitted 29 November, 2024; v1 submitted 1 May, 2024; originally announced May 2024.

AttackBench: Evaluating Gradient-based Attacks for Adversarial Examples

Authors: Antonio Emanuele Cinà, Jérôme Rony, Maura Pintor, Luca Demetrio, Ambra Demontis, Battista Biggio, Ismail Ben Ayed, Fabio Roli

Abstract: Adversarial examples are typically optimized with gradient-based attacks. While novel attacks are continuously proposed, each is shown to outperform its predecessors using different experimental setups, hyperparameter settings, and number of forward and backward calls to the target models. This provides overly-optimistic and even biased evaluations that may unfairly favor one particular attack ove… ▽ More Adversarial examples are typically optimized with gradient-based attacks. While novel attacks are continuously proposed, each is shown to outperform its predecessors using different experimental setups, hyperparameter settings, and number of forward and backward calls to the target models. This provides overly-optimistic and even biased evaluations that may unfairly favor one particular attack over the others. In this work, we aim to overcome these limitations by proposing AttackBench, i.e., the first evaluation framework that enables a fair comparison among different attacks. To this end, we first propose a categorization of gradient-based attacks, identifying their main components and differences. We then introduce our framework, which evaluates their effectiveness and efficiency. We measure these characteristics by (i) defining an optimality metric that quantifies how close an attack is to the optimal solution, and (ii) limiting the number of forward and backward queries to the model, such that all attacks are compared within a given maximum query budget. Our extensive experimental analysis compares more than $100$ attack implementations with a total of over $800$ different configurations against CIFAR-10 and ImageNet models, highlighting that only very few attacks outperform all the competing approaches. Within this analysis, we shed light on several implementation issues that prevent many attacks from finding better solutions or running at all. We release AttackBench as a publicly-available benchmark, aiming to continuously update it to include and evaluate novel gradient-based attacks for optimizing adversarial examples. △ Less

Submitted 12 May, 2025; v1 submitted 30 April, 2024; originally announced April 2024.

NC-TTT: A Noise Contrastive Approach for Test-Time Training

Authors: David Osowiechi, Gustavo A. Vargas Hakim, Mehrdad Noori, Milad Cheraghalikhani, Ali Bahri, Moslem Yazdanpanah, Ismail Ben Ayed, Christian Desrosiers

Abstract: Despite their exceptional performance in vision tasks, deep learning models often struggle when faced with domain shifts during testing. Test-Time Training (TTT) methods have recently gained popularity by their ability to enhance the robustness of models through the addition of an auxiliary objective that is jointly optimized with the main task. Being strictly unsupervised, this auxiliary objectiv… ▽ More Despite their exceptional performance in vision tasks, deep learning models often struggle when faced with domain shifts during testing. Test-Time Training (TTT) methods have recently gained popularity by their ability to enhance the robustness of models through the addition of an auxiliary objective that is jointly optimized with the main task. Being strictly unsupervised, this auxiliary objective is used at test time to adapt the model without any access to labels. In this work, we propose Noise-Contrastive Test-Time Training (NC-TTT), a novel unsupervised TTT technique based on the discrimination of noisy feature maps. By learning to classify noisy views of projected feature maps, and then adapting the model accordingly on new domains, classification performance can be recovered by an important margin. Experiments on several popular test-time adaptation baselines demonstrate the advantages of our method compared to recent approaches for this task. The code can be found at:https://github.com/GustavoVargasHakim/NCTTT.git △ Less

Submitted 12 April, 2024; originally announced April 2024.