arXiv subjects

Hilde Kuehne

Publications and source records attributed to Hilde Kuehne.

At least 19 recordsLinked to original sources

arXiv2026

TeMo: Temperature Modulation for Multimodal Contrastive Learning

Contrastive learning approaches achieve strong performance by training models to bring similar samples closer while pushing dissimilar samples apart. A crucial component of contrastive learning is the temperature hyperparameter $\tau$, which controls the penalty strength applied to negative samples. However, most existing methods either fix this hyperparameter or learn a global value during training. In this paper, we introduce TeMo, Temperature Modulation framework, a similarity-based modulation approach that adaptively adjusts the temperature for each positive-negative pair according to their similarity, enabling more fine-grained multimodal contrastive learning. Our approach seamlessly integrates temperature-modulated multimodal and unimodal losses with the standard multimodal contrastive loss by gradually transitioning between them. This design allows the model to capture both coarse- and fine-grained semantics at different training stages. Extensive experiments demonstrate that each component of TeMo consistently enhances performance across diverse zero-shot retrieval and classification tasks, establishing new state-of-the-art results.

cs.CV↗

arXiv2026

Falcon Perception-HD: High Density Perception via Reinforcement Learning

Autoregressive perception models trained to localize visual entities under the open-vocabulary setting are mostly trained using Supervised fine-tuning (SFT) with maximum likelihood, yet it optimizes a proxy objective (per-token cross-entropy) that is fundamentally misaligned with perception metrics such as precision and recall. In this paper, we explore post-training reinforcement learning (RL), specifically GRPO, to directly align these models with their evaluation metrics. Building up on the recently introduced Falcon Perception, we design an RL framework that addresses perception-specific challenges: reward design for set-structured outputs and multi-head sampling control. We discover multiple benefits from RL for perception: first, RL unlocks state-of-the-art performance in very dense scenes (up to 500 objects per scene), a regime where most existing systems degrade sharply or collapse; furthermore it fixes common issues in autoregressive perception models like mask repetitions and removes almost entirely the need for NMS and coordinate deduplication, which improve both performance and efficiency and remove the need for hyperparameters tuning; overall, we notice improvements on all levels of difficulties in referring expression segmentation (on PBench and SACO-Gold), and we find an elegant way to preserve the knowledge of whether an object exists or not (as evaluated by MCC) without training on negative samples. We show that a simple reward that penalizes false negatives and positives is sufficient. We develop two hybrid self-annotation pipelines, respectively tailored for difficult referring expressions and very dense scenes, and show their benefits on RL-training. Model weights are released as a Falcon Perception revision~\footnote{https://huggingface.co/tiiuae/Falcon-Perception}. Datasets will be published.

cs.CV↗

arXiv2026

Decompose, Compare, and Decide: Multimodal LLMs are Implicit Few-Shot Learners

Multimodal Large Language Models (MLLMs) have demonstrated remarkable abilities when analyzing images, yet translating these capabilities to few-shot image classification remains challenging. To bridge this gap, we present DeCoDe, a simple yet effective technique that enables off-the-shelf MLLMs to act as strong few-shot classifiers without any additional training. Our approach builds on the idea of few-shot classification as a set of pairwise image comparisons, decomposing the task into a set of binary decisions. Given a query image and a support image from a candidate class, the MLLM is prompted to decide whether the two images depict the same class. The logit corresponding to an affirmative response is then used as a similarity score to assign the query image to the most likely class. While this already yields good results, we show that providing additional high-level information, such as the data domain, to the model further improves performance. Our evaluation provides an extensive analysis of various inference variants on a suite of twelve datasets, six established and six newly curated few-shot benchmarks spanning across diverse domains. The results show that the proposed simple decomposition technique can turn off-the-shelf MLLMs into powerful few-shot learners, significantly outperforming current state-of-the-art few-shot methods on both standard and novel domains. Code is available at https://github.com/yunhanwang1105/DeCoDe.

cs.CV↗

arXiv2026

DataComp-VLM: Improved Open Datasets for Vision-Language Models

Building performant Vision-Language Models (VLMs) requires carefully curating large-scale training datasets, yet the community lacks systematic benchmarks for evaluating such curation strategies. We introduce DataComp for VLMs (DCVLM), a benchmark for controlled data-centric experiments to improve VLM training. As part of DCVLM, we collect 160 datasets spanning four data types -- image-caption pairs, multimodal interleaved documents, text-only, and instruction-tuning data -- into a corpus of 6T multimodal tokens. DCVLM allows participants to test curation strategies (filtering, mixing, formatting, sampling) across 1B-8B models and 6.25B-200B token budgets. Models are then evaluated on a carefully selected suite of up to 52 downstream benchmarks across 9 domains. We conduct extensive experiments on DCVLM and find that data mixing, not filtering, is key to a high-quality training dataset: instruction-heavy mixtures scale better than caption-heavy ones, with gains widening at larger scales. The resulting dataset, DCVLM-Baseline, enables training an 8B VLM to 63.6% accuracy on our 33-task core suite with 200B training tokens. Compared to FineVision, the state-of-the-art open VLM training dataset, this represents an improvement of +5.4pp. DCVLM and all accompanying artifacts will be made publicly available at https://www.datacomp.ai/dcvlm/.

cs.CV↗

arXiv2026

Physics-IQ Verified

Video generative models ( VGMs) have become a new frontier that can be used not just for video generation but for a multitude of downstream tasks, including world modeling. To advance these tasks, a good video model must understand the physical reality of the world. Evaluating this understanding is an emerging field and has led to the Physics-IQ benchmark, which quantifies this explicitly by comparing model-generated videos to real-world videos of physical experiments. In this work, we present a systematic audit of the Physics-IQ benchmark, expose shortcomings and propose three solutions that sharpen how we can measure physical understanding of VGMs. Specifically, we improve prompt and ground-truth quality to reduce the influence of confounding factors and further introduce a sample-level scoring system that weights each sample and metric equally. Our resulting benchmark, Physics-IQ Verified, refines 57.6\% of all samples and improves over 34.8\% of prompts. In a comparison study using six image-to-video generative models, we observe moderate but meaningful ranking changes (Kendall's $τ= 0.46$). We hope Physics-IQ Verified advances the community by providing a more reliable signal toward physically accurate VGMs. The code for the benchmark can be accessed at https://github.com/google-deepmind/physics-iq-benchmark

cs.CV↗

arXiv2026

VOLD: Reasoning Transfer from LLMs to Vision-Language Models via On-Policy Distillation

Training vision-language models (VLMs) for complex reasoning remains a challenging task, i.a. due to the scarcity of high-quality image-text reasoning data. Conversely, text-based reasoning resources are abundant and scalable, but it is still an open question how to leveraging them for VLM reasoning. To address this problem, we propose VOLD, a framework to transfer reasoning capabilities from text-only teacher models to VLM student models. To this end, VOLD combines reinforcement learning via Group Relative Policy Optimization (GRPO) with on-policy distillation, which allows the student reasoning traces to be guided by the teacher model, resulting in a significant gain over using GRPO alone. We further show that a cold-start alignment is essential for an effective transfer during the online training phase in this scenario and that without sufficient distributional alignment between teacher and student, on-policy distillation fails to provide meaningful guidance. We evaluate VOLD across diverse benchmarks including MMMU-Pro, MathVision, MathVista, and LogicVista, showing that VOLD outperforms the baseline model significantly and improves over the state of the art by a margin. Our ablation shows the importance of a cold-start alignment via SFT for on-policy distillation with a text-only teacher.

cs.CV↗

arXiv2026

FindIt: A Format-Informed Visual Detection Benchmark for Generalist Multimodal LLMs

Multimodal large language models (MLLMs) are predominantly evaluated on free-form vision-language tasks such as visual question answering, captioning, and summarization. However, their practical use is rapidly expanding to more structured computer vision settings, where users prompt models to perform localization-centric tasks such as object detection, often within larger agentic or decision-making systems. Despite this shift, there is currently no standardized benchmark that systematically evaluates these capabilities at scale. In this work, we introduce the first comprehensive benchmark specifically designed to assess the promptable localization abilities of generalist MLLMs. Our benchmark spans four core task categories: object detection, referring expression detection, instance-level detection, and video-based detection. To enable consistent and fair evaluation, we develop a unified framework that standardizes inputs, enforces parsable bounding box outputs, and defines transparent evaluation protocols across tasks. Using this suite, we evaluate a diverse set of open-source and proprietary MLLMs, providing an in-depth analysis of their performance and limitations. Beyond accuracy, we examine models' ability to adhere to output format specifications, showing that current systems are highly sensitive to formatting constraints and often fail to generalize even to minor variations. Our results highlight both the strengths and shortcomings of state-of-the-art MLLMs in localization settings, and point toward important directions for improving multimodal model design and evaluation.

cs.CV↗

arXiv2026

VisualOverload: Probing Visual Understanding of VLMs in Really Dense Scenes

Is basic visual understanding really solved in state-of-the-art VLMs? We present VisualOverload, a slightly different visual question answering (VQA) benchmark comprising 2,720 question-answer pairs, with privately held ground-truth responses. Unlike prior VQA datasets that typically focus on near global image understanding, VisualOverload challenges models to perform simple, knowledge-free vision tasks in densely populated (or, overloaded) scenes. Our dataset consists of high-resolution scans of public-domain paintings that are populated with multiple figures, actions, and unfolding subplots set against elaborately detailed backdrops. We manually annotated these images with questions across six task categories to probe for a thorough understanding of the scene. We hypothesize that current benchmarks overestimate the performance of VLMs, and encoding and reasoning over details is still a challenging task for them, especially if they are confronted with densely populated scenes. Indeed, we observe that even the best model (o3) out of 37 tested models only achieves 19.6% accuracy on our hardest test split and overall 69.5% accuracy on all questions. Beyond a thorough evaluation, we complement our benchmark with an error analysis that reveals multiple failure modes, including a lack of counting skills, failure in OCR, and striking logical inconsistencies under complex tasks. Altogether, VisualOverload exposes a critical gap in current vision models and offers a crucial resource for the community to develop better models. Benchmark: http://paulgavrikov.github.io/visualoverload

cs.CV↗

arXiv2026

AVRT: Audio-Visual Reasoning Transfer through Single-Modality Teachers

Recent advances in reasoning models have shown remarkable progress in text-based domains, but transferring those capabilities to multimodal settings, e.g., to allow reasoning over audio-visual data, still remains a challenge, in part because of the limited availability of high-quality reasoning data in targeted multimodal combinations. To address this problem, we introduce AVRT, a novel framework that generates high-quality audio-visual reasoning traces from single-modality teacher models. We generate independent vision- and audio-reasoning traces via models specialized to reason over their respective modalities and merge the resulting traces with an LLM merger model. The resulting multimodal traces are used in a supervised fine-tuning (SFT) cold start to adapt the target model to audio-visual reasoning traces first, before training it in a second reinforcement learning stage on larger-scale data. Evaluated on seven audio-visual and audio benchmarks, our 3B and 7B parameter models achieve state-of-the-art results among models of comparable size including OmniBench and DailyOmni for audio-visual and MMAR for audio-only reasoning, showing that cross-modal training also transfers to single-modality tasks and establishing a new training pipeline for multimodal reasoning models.

cs.CV↗

arXiv2026

SigLino: Efficient Multi-Teacher Distillation for Agglomerative Vision Foundation Models

Vision foundation models trained via multi-teacher distillation offer a promising path toward unified visual representations, yet the learning dynamics and data efficiency of such approaches remain underexplored. In this paper, we systematically study multi-teacher distillation for vision foundation models and identify key factors that enable training at lower computational cost. We introduce SigLino, an efficient family of agglomerative vision foundation models that distill knowledge from SigLIP2 and DINOv3 simultaneously into Dense and Mixture-of-Experts students. We show that (1) our Asymmetric Relation-Knowledge Distillation loss preserves the geometric properties of each teacher while enabling effective knowledge transfer, (2) token-balanced batching that packs varying-resolution images into sequences with uniform token budgets stabilizes representation learning across resolutions without sacrificing performance, (3) hierarchical clustering and sampling of training data, typically reserved for self-supervised learning, substantially improves sample efficiency over random sampling for multi-teacher distillation, and (4) the resulting representations transfer effectively to early-fusion Grounding-VLMs, outperforming models trained from scratch. By combining these findings, we curate OpenLVD200M, a 200M-image corpus that demonstrates superior efficiency for multi-teacher distillation. Instantiated in a Mixture-of-Experts, our SigLino-MoE initializes an early-fusion Grounding-VLM that replaces the conventional ViT->LLM stack, demonstrating improved performance compared to a model trained from scratch. We release OpenLVD200M and five distilled checkpoints comprising MoE and dense variants.

cs.CV↗

arXiv2026

TTA-Vid: Generalized Test-Time Adaptation for Video Reasoning

Recent video reasoning models have shown strong results on temporal and multimodal understanding, yet they depend on large-scale supervised data and multi-stage training pipelines, making them costly to train and difficult to adapt to new domains. In this work, we leverage the paradigm of Test-Time Reinforcement Learning on video-language data to allow for adapting a pretrained model to incoming video samples at test-time without explicit labels. The proposed test-time adaptation for video approach (TTA-Vid) combines two components that work simultaneously: (1) a test-time adaptation that performs step-by-step reasoning at inference time on multiple frame subsets. We then use a batch-aware frequency-based reward computed across different frame subsets as pseudo ground truth to update the model. It shows that the resulting model trained on a single batch or even a single sample from a dataset, is able to generalize at test-time to the whole dataset and even across datasets. Because the adaptation occurs entirely at test time, our method requires no ground-truth annotations or dedicated training splits. Additionally, we propose a multi-armed bandit strategy for adaptive frame selection that learns to prioritize informative frames, guided by the same reward formulation. Our evaluation shows that TTA-Vid yields consistent improvements across various video reasoning tasks and is able to outperform current state-of-the-art methods trained on large-scale data. This highlights the potential of test-time reinforcement learning for temporal multimodal understanding.

cs.CV↗

arXiv2026

CALM: Class-Conditional Sparse Attention Vectors for Large Audio-Language Models

Large audio-language models (LALMs) exhibit strong zero-shot capabilities in multiple downstream tasks, such as audio question answering (AQA) and abstract reasoning; however, these models still lag behind specialized models for certain discriminative tasks (e.g., audio classification). Recent studies show that sparse subsets of attention heads within an LALM can serve as strong discriminative feature extractors for downstream tasks such as classification via simple voting schemes. However, these methods assign uniform weights to all selected heads, implicitly assuming that each head contributes equally across all semantic categories. In this work, we propose Class-Conditional Sparse Attention Vectors for Large Audio-Language Models, a few-shot classification method that learns class-dependent importance weights over attention heads. This formulation allows individual heads to specialize in distinct semantic categories and to contribute to ensemble predictions proportionally to their estimated reliability. Experiments on multiple few-shot audio and audiovisual classification benchmarks and tasks demonstrate that our method consistently outperforms state-of-the-art uniform voting-based approaches by up to 14.52%, 1.53%, 8.35% absolute gains for audio classification, audio-visual classification, and spoofing detection respectively.

cs.SD↗

arXiv2026

MM-TS: Multi-Modal Temperature and Margin Schedules for Contrastive Learning with Long-Tail Data

Contrastive learning has become a fundamental approach in both uni-modal and multi-modal frameworks. This learning paradigm pulls positive pairs of samples closer while pushing negatives apart. In the uni-modal setting (e.g., image-based learning), previous research has shown that the strength of these forces can be controlled through the temperature parameter. In this work, we propose Multi-Modal Temperature and Margin Schedules (MM-TS), extending the concept of uni-modal temperature scheduling to multi-modal contrastive learning. Our method dynamically adjusts the temperature in the contrastive loss during training, modulating the attraction and repulsion forces in the multi-modal setting. Additionally, recognizing that standard multi-modal datasets often follow imbalanced, long-tail distributions, we adapt the temperature based on the local distribution of each training sample. Specifically, samples from dense clusters are assigned a higher temperature to better preserve their semantic structure. Furthermore, we demonstrate that temperature scheduling can be effectively integrated within a max-margin framework, thereby unifying the two predominant approaches in multi-modal contrastive learning: InfoNCE loss and max-margin objective. We evaluate our approach on four widely used image- and video-language datasets, Flickr30K, MSCOCO, EPIC-KITCHENS-100, and YouCook2, and show that our dynamic temperature and margin schedules improve performance and lead to new state-of-the-art results in the field.

cs.CV↗

arXiv2026

DEX-AR: A Dynamic Explainability Method for Autoregressive Vision-Language Models

As Vision-Language Models (VLMs) become increasingly sophisticated and widely used, it becomes more and more crucial to understand their decision-making process. Traditional explainability methods, designed for classification tasks, struggle with modern autoregressive VLMs due to their complex token-by-token generation process and intricate interactions between visual and textual modalities. We present DEX-AR (Dynamic Explainability for AutoRegressive models), a novel explainability method designed to address these challenges by generating both per-token and sequence-level 2D heatmaps highlighting image regions crucial for the model's textual responses. The proposed method offers to interpret autoregressive VLMs-including varying importance of layers and generated tokens-by computing layer-wise gradients with respect to attention maps during the token-by-token generation process. DEX-AR introduces two key innovations: a dynamic head filtering mechanism that identifies attention heads focused on visual information, and a sequence-level filtering approach that aggregates per-token explanations while distinguishing between visually-grounded and purely linguistic tokens. Our evaluation on ImageNet, VQAv2, and PascalVOC, shows a consistent improvement in both perturbation-based metrics, using a novel normalized perplexity measure, as well as segmentation-based metrics.

cs.CV↗

arXiv2026

MaskInversion: Localized Embeddings via Optimization of Explainability Maps

Vision-language foundation models such as CLIP have achieved tremendous results in global vision-language alignment, but still show some limitations in creating representations for specific image regions. % To address this problem, we propose MaskInversion, a method that leverages the feature representations of pre-trained foundation models, such as CLIP, to generate a context-aware embedding for a query image region specified by a mask at test time. MaskInversion starts with initializing an embedding token and compares its explainability map, derived from the foundation model, to the query mask. The embedding token is then subsequently refined to approximate the query region by minimizing the discrepancy between its explainability map and the query mask. During this process, only the embedding vector is updated, while the underlying foundation model is kept frozen allowing to use MaskInversion with any pre-trained model. As deriving the explainability map involves computing its gradient, which can be expensive, we propose a gradient decomposition strategy that simplifies this computation. The learned region representation can be used for a broad range of tasks, including open-vocabulary class retrieval, referring expression comprehension, as well as for localized captioning and image generation. We evaluate the proposed method on all those tasks on several datasets such as PascalVOC, MSCOCO, RefCOCO, and OpenImagesV7 and show its capabilities compared to other SOTA approaches.

cs.CV↗

arXiv2026

When LLaVA Meets Objects: Token Composition for Vision-Language-Models

Current autoregressive Vision Language Models (VLMs) usually rely on a large number of visual tokens to represent images, resulting in a need for more compute especially at inference time. To address this problem, we propose Mask-LLaVA, a framework that leverages different levels of visual features to create a compact yet information-rich visual representation for autoregressive VLMs. Namely, we combine mask-based object representations together with global tokens and local patch tokens. While all tokens are used during training, it shows that the resulting model can flexibly drop especially the number of mask-based object-tokens at test time, allowing to adapt the number of tokens during inference without the need to retrain the model and without a significant drop in performance. We evaluate the proposed approach on a suite of standard benchmarks showing results competitive to current token efficient methods and comparable to the original LLaVA baseline using only a fraction of visual tokens. Our analysis demonstrates that combining multi-level features enables efficient learning with fewer tokens while allowing dynamic token selection at test time for good performance.

cs.CV↗

arXiv2025

TTRV: Test-Time Reinforcement Learning for Vision Language Models

Existing methods for extracting reward signals in Reinforcement Learning typically rely on labeled data and dedicated training splits, a setup that contrasts with how humans learn directly from their environment. In this work, we propose TTRV to enhance vision language understanding by adapting the model on the fly at inference time, without the need for any labeled data. Concretely, we enhance the Group Relative Policy Optimization (GRPO) framework by designing rewards based on the frequency of the base model's output, while inferring on each test sample multiple times. Further, we also propose to control the diversity of the model's output by simultaneously rewarding the model for obtaining low entropy of the output empirical distribution. Our approach delivers consistent gains across both object recognition and visual question answering (VQA), with improvements of up to 52.4% and 29.8%, respectively, and average boosts of 24.6% and 10.0% across 16 datasets. Remarkably, on image recognition, TTRV applied to InternVL 8B surpasses GPT-4o by an average of 2.3% over 8 benchmarks, while remaining highly competitive on VQA, demonstrating that test-time reinforcement learning can match or exceed the strongest proprietary models. Finally, we find many interesting properties of test-time RL for VLMs: for example, even in extremely data-constrained scenarios, where adaptation is performed on a single randomly chosen unlabeled test example, TTRV still yields non-trivial improvements of up to 5.5% in recognition tasks.

cs.CV↗

arXiv2025

Towards Audio Token Compression in Large Audio Language Models

Large Audio Language Models (LALMs) deliver strong performance across speech and audio tasks, but their audio encoders generate high-rate token sequences (e.g., 25 tokens/s), making attention computation costly and limiting scalability. In this paper, we explore techniques such as unsupervised segmentation, uniform average pooling, etc., to reduce the number of audio tokens before they are consumed by the LLM decoder. To mitigate potential performance degradation, we employ low-rank adapters during finetuning. We evaluate our proposed models on two tasks, automatic speech recognition and speech-to-speech translation tasks, that are dependent on effectively uncovering the underlying lexical content of the input signal, and study the effect of downsampling on these tasks. Experimental results show that compressed LALMs can achieve performance closer to frame-level LALMs while reducing the input audio token count up to three times before the LLM backbone.

eess.AS↗