In deep learning, functional similarity evaluation quantifies the extent to which independently trained models learn similar input--output relationships. In model stitching, functional similarity is framed as representation forward compatibility, i.e., whether the representations of two models can be aligned to solve a given task. Recent studies, however, highlight a critical limitation: models relying on different information cues can still produce compatible representations, making them appear misleadingly similar (Smith et al., 2025). We attribute this failure to standard model stitching being inherently blind to the invariance properties of the stitched models. To address this limitation, we introduce the forward--backward compatibility requirement under which we formulate the invariance-aware model stitching. Through analyzing key stitching configurations, we study the interplay between forward and backward compatibility, showing that invariance-aware model stitching provides a more principled approach to functional similarity evaluation while revealing functional discrepancies previously obscured.

Large multimodal models (LMMs) have recently gained attention due to their effectiveness to understand and generate descriptions of visual content. Most existing LMMs are in English language. While few recent works explore multilingual image LMMs, to the best of our knowledge, moving beyond the English language for cultural and linguistic inclusivity is yet to be investigated in the context of video LMMs. In pursuit of more inclusive video LMMs, we introduce a multilingual Video LMM benchmark, named ViMUL-Bench, to evaluate Video LMMs across 14 languages, including both low- and high-resource languages: Arabic, Bengali, Chinese, English, French, German, Hindi, Japanese, Russian, Sinhala, Spanish, Swedish, Tamil, and Urdu. Our ViMUL-Bench is designed to rigorously test video LMMs across 15 categories including eight culturally diverse categories, ranging from lifestyles and festivals to foods and rituals and from local landmarks to prominent cultural personalities. ViMUL-Bench comprises both open-ended (short and long-form) and multiple-choice questions spanning various video durations (short, medium, and long) with 8k samples that are manually verified by native language speakers. In addition, we also introduce a machine translated multilingual video training set comprising 1.2 million samples and develop a simple multilingual video LMM, named ViMUL, that is shown to provide a better tradeoff between high-and low-resource languages for video understanding. We hope our ViMUL-Bench and multilingual video LMM along with a large-scale multilingual video training set will help ease future research in developing cultural and linguistic inclusive multilingual video LMMs. Our proposed benchmark, video LMM and training data will be publicly released.

Variational Autoencoders (VAEs) are a popular framework for unsupervised learning and data generation. A plethora of methods have been proposed focusing on improving VAEs,with the incorporation of adversarial objectives and the integration of prior learning mechanismsbeing prominent directions. When it comes to the former, an indicative instance is therecently introduced family of Introspective VAEs aiming at ensuring that a low likelihood isassigned to unrealistic samples. In this study, we focus on the Soft-IntroVAE (S-IntroVAE),one of only two members of the Introspective VAE family, the other being the originalIntroVAE. We select S-IntroVAE for its state-of-the-art status and its training stability.In particular, we investigate the implication of incorporating a multimodal and trainableprior into this S-IntroVAE. Namely, we formulate the prior as a third player and show thatwhen trained in cooperation with the decoder constitutes an effective way for prior learning,which shares the Nash Equilibrium with the vanilla S-IntroVAE. Furthermore, basedon a modified formulation of the optimal ELBO in S-IntroVAE, we develop theoreticallymotivated regularizations, namely (i) adaptive variance clipping to stabilize training whenlearning the prior and (ii) responsibility regularization to discourage the formation of inactiveprior modes. Finally, we perform a series of targeted experiments on a 2D densityestimation benchmark and in an image generation setting comprised of the (F)-MNIST andCIFAR-10 datasets demonstrating the effect of prior learning in S-IntroVAE in generationand representation learning.

Coverage path planning (CPP) is the problem of finding a path that covers the entire free space of a confined area, with applications ranging from robotic lawn mowing to search-and-rescue. While for known environments, offline methods can find provably complete paths, and in some cases optimal solutions, unknown environments need to be planned online during mapping. We investigate the suitability of continuous-space reinforcement learning (RL) for this challenging problem, and propose a computationally feasible egocentric map representation based on frontiers, as well as a novel reward term based on total variation to promote complete coverage. Compared to existing classical methods, this approach allows for a flexible path space, and enables the agent to adapt to specific environment characteristics. Meanwhile, the deployment of RL models on real robot systems is difficult. Training from scratch may be infeasible due to slow convergence times, while transferring from simulation to reality, i.e. sim-to-real transfer, is a key challenge in itself. We bridge the sim-to-real gap through a semi-virtual environment, including a real robot and real-time aspects, while utilizing a simulated sensor and obstacles to enable environment randomization and automated episode resetting. We investigate what level of fine-tuning is needed for adapting to a realistic setting. Through extensive experiments, we show that our approach surpasses the performance of both previous RL-based approaches and highly specialized methods across multiple CPP variations in simulation. Meanwhile, our method successfully transfers to a real robot. Our code implementation can be found online (Link to code repository: https://github.com/arvijj/rl-cpp).

The VOTS2025 challenge marks the thirteenth edition of the Visual Object Tracking Segmentation benchmarking activity organized under the VOT initiative. Building on the tracking setup introduced in VOTS2023, the challenge continues to integrate short-term and long-term tracking, as well as single-target and multi-target scenarios, using segmentation masks as the sole form of target annotation. This year's benchmark features three sub-challenges. The first two, VOTS2025 and VOTSt2025, evaluate tracking of conventional objects and objects undergoing topological changes, respectively. A new addition, VOTS-RT2025, aims to foster the development of efficient tracking models by introducing constraints that highlight realtime performance. All sub-challenges adopt a consistent evaluation protocol, with VOTS-RT2025 introducing specific modifications to reflect latency-aware performance. We report and analyze results from 32 submissions. Full tracker descriptions, source code, datasets, and the evaluation toolkit are available on the project website https://www.votchallenge.net/vots2025.

We propose a novel theoretical framework that unifies geometric and algebraic perspectives on uncertainty quantification in deep generative models. Standard variational autoencoders (VAEs) often underestimate uncertainty in the presence of non-commutative transformations, leading to miscalibrated confidence and potential fairness violations. Our approach introduces an integrated diagnostic and regularization framework that monitors transformation relationships via Baker--Campbell--Hausdorff deviation and decoder order-swap tests; then, we apply category-specific regularization. Commutative pairs receive standard penalties, decoder-induced artifacts are suppressed through equivariance constraints, and genuinely non-commutative transformations are governed by a deformation-stability principle linking commutator strength to required uncertainty scaling. The key theoretical result establishes that geometric uncertainty, measured through the Riemannian sensitivity of the decoder to Lie-algebra actions, should scale with algebraic non-commutativity to ensure proper calibration. This framework closes a gap in generative modelling by providing principled diagnostics and regularization strategies for geometrically-aware uncertainty in symmetry-rich latent spaces, with direct implications for fairness when biased correlations induce spurious non-commutativity.

Keypoint detection is a pivotal step in 3D reconstruction, whereby sets of (up to) K points are detected in each view of a scene. Crucially, the detected points need to be consistent between views, i.e., correspond to the same 3D point in the scene. One of the main challenges with keypoint detection is the formulation of the learning objective. Previous learning-based methods typically jointly learn descriptors with keypoints, and treat the keypoint detection as a binary classification task on mutual nearest neighbours. However, basing keypoint detection on descriptor nearest neighbours is a proxy task, which is not guaranteed to produce 3D-consistent keypoints. Furthermore, this ties the keypoints to a specific descriptor, complicating downstream usage. In this work, we instead learn keypoints directly from 3D consistency. To this end, we train the detector to detect tracks from large-scale SfM. As these points are often overly sparse, we derive a semi-supervised two-view detection objective to expand this set to a desired number of detections. To train a descriptor, we maximize the mutual nearest neighbour objective over the keypoints with a separate network. Results show that our approach, DeDoDe, achieves significant gains on multiple geometry benchmarks. Code is provided at http://github.com/Parskatt/DeDoDegithub.com/Parskatt/DeDoDe.

Long-tailed semi-supervised learning (LTSSL) represents a practical scenario for semi-supervised applications, challenged by skewed labeled distributions that bias classifiers. This problem is often aggravated by discrepancies between labeled and unlabeled class distributions, leading to biased pseudo-labels, neglect of rare classes, and poorly calibrated probabilities. To address these issues, we introduce Flexible Distribution Alignment (FlexDA), a novel adaptive logit-adjusted loss framework designed to dynamically estimate and align predictions with the actual distribution of unlabeled data and achieve a balanced classifier by the end of training. FlexDA is further enhanced by a distillation-based consistency loss, promoting fair data usage across classes and effectively leveraging underconfident samples. This method, encapsulated in ADELLO (Align and Distill Everything All at Once), proves robust against label shift, significantly improves model calibration in LTSSL contexts, and surpasses previous state-of-of-art approaches across multiple benchmarks, including CIFAR100-LT, STL10-LT, and ImageNet127, addressing class imbalance challenges in semi-supervised learning. Our code is available at https://github.com/emasa/ADELLO-LTSSL.

Image-level weakly-supervised semantic segmentation (WSSS) reduces the usually vast data annotation cost by surrogate segmentation masks during training. The typical approach involves training an image classification network using global average pooling (GAP) on convolutional feature maps. This enables the estimation of object locations based on class activation maps (CAMs), which identify the importance of image regions. The CAMs are then used to generate pseudo-labels, in the form of segmentation masks, to supervise a segmentation model in the absence of pixel-level ground truth. Our work is based on two techniques for improving CAMs; importance sampling, which is a substitute for GAP, and the feature similarity loss, which utilizes a heuristic that object contours almost always align with color edges in images. However, both are based on the multinomial posterior with softmax, and implicitly assume that classes are mutually exclusive, which turns out suboptimal in our experiments. Thus, we reformulate both techniques based on binomial posteriors of multiple independent binary problems. This has two benefits; their performance is improved and they become more general, resulting in an add-on method that can boost virtually any WSSS method. This is demonstrated on a wide variety of baselines on the PASCAL VOC dataset, improving the region similarity and contour quality of all implemented state-of-the-art methods. Experiments on the MS COCO dataset further show that our proposed add-on is well-suited for large-scale settings. Our code implementation is available at https://github.com/arvijj/hfpl.

Coverage path planning (CPP) is the problem of finding a path that covers the entire free space of a confined area, with applications ranging from robotic lawn mowing to search-and-rescue. When the environment is unknown, the path needs to be planned online while mapping the environment, which cannot be addressed by offline planning methods that do not allow for a flexible path space. We investigate how suitable reinforcement learning is for this challenging problem, and analyze the involved components required to efficiently learn coverage paths, such as action space, input feature representation, neural network architecture, and reward function. We propose a computationally feasible egocentric map representation based on frontiers, and a novel reward term based on total variation to promote complete coverage. Through extensive experiments, we show that our approach surpasses the performance of both previous RL-based approaches and highly specialized methods across multiple CPP variations.