Abstract Large Language Models (LLMs) exhibit In-Context Learning (ICL), which enables the model to perform new tasks conditioning only on the examples provided in the context without updating the model’s weights. While ICL offers fast adaptation across natural language tasks and domains, its emergence is less straightforward for modalities beyond text. In this work, we systematically uncover properties present in LLMs that support the emergence of ICL for autoregressive models and various modalities by promoting the learning of the mechanisms needed for ICL. We identify exact token repetitions in the training data sequences as an important factor for ICL. Such repetitions further improve stability and reduce transiency in ICL performance. We analyse in detail the training dynamics of such data sequences and explain how token repetitions enhance the ICL learning mechanisms. Moreover, we emphasise the importance of the training task difficulty for the emergence of ICL. Finally, by applying our novel insights on ICL emergence, we unlock ICL capabilities across various visual datasets used for few-shot classification, and confirm the generalisability of our insights to much harder real-world examples of large-scale object classification, and a more challenging EEG classification task. Code is available at https://github.com/jelenab98/unlocking_icl

Existing world models for autonomous driving struggle with long-horizon generation and generalization to challenging scenarios. In this work, we develop a model using simple design choices, and without additional supervision or sensors, such as maps, depth, or multiple cameras. We show that our model yields state-of-the-art performance, despite having only 469M parameters and being trained on 280h of video data. It particularly stands out in difficult scenarios like turning maneuvers and urban traffic. We test whether discrete token models possibly have advantages over continuous models based on flow matching. To this end, we set up a hybrid tokenizer that is compatible with both approaches and allows for a side-by-side comparison. Our study concludes in favor of the continuous autoregressive model, which is less brittle on individual design choices and more powerful than the model built on discrete tokens. Project page with code, model checkpoints and visualization can be found here: https://lmb-freiburg.github.io/orbis.github.io/

Abstract Active learning aims to reduce the high labeling cost involved in training machine learning models on large datasets by efficiently labeling only the most informative samples. Recently, deep active learning has shown success on various tasks. However, the conventional evaluation schemes are either incomplete or below par. This study critically assesses various active learning approaches, identifying key factors essential for choosing the most effective active learning method. It includes a comprehensive guide to obtain the best performance for each case, in image classification and semantic segmentation. For image classification, the AL methods improve by a large-margin when integrated with data augmentation and semi-supervised learning, but barely perform better than the random baseline. In this work, we evaluate them under more realistic settings and propose a more suitable evaluation protocol. For semantic segmentation, previous academic studies focused on diverse datasets with substantial annotation resources. In contrast, data collected in many driving scenarios is highly redundant, and most medical applications are subject to very constrained annotation budgets. The study evaluates active learning techniques under various conditions including data redundancy, the use of semi-supervised learning, and differing annotation budgets. As an outcome of our study, we provide a comprehensive usage guide to obtain the best performance for each case.

Vision-language modeling has enabled open-vocabulary tasks where predictions can be queried using any text prompt in a zero-shot manner. Existing open-vocabulary tasks focus on object classes, whereas research on object attributes is limited due to the lack of a reliable attribute-focused evaluation benchmark. This paper introduces the Open- Vocabulary Attribute Detection (OVAD) task and the corresponding OVAD benchmark. The objective of the novel task and benchmark is to probe object-level attribute information learned by vision-language models. To this end, we created a clean and densely annotated test set covering 117 attribute classes on the 80 object classes of MS COCO. It includes positive and negative annotations, which enables open-vocabulary evaluation. Overall, the benchmark consists of 1.4 million annotations. For reference, we provide a first baseline method for open-vocabulary attribute detection. Moreover, we demonstrate the benchmark's value by studying the attribute detection performance of several foundation models.

The ability to understand visual information from limited labeled data is an important aspect of machine learning. While image-level classification has been extensively studied in a semi-supervised setting, dense pixel-level classification with limited data has only drawn attention recently. In this work, we propose an approach for semi-supervised semantic segmentation that learns from limited pixel-wise annotated samples while exploiting additional annotation-free images. The proposed approach relies on adversarial training with a feature matching loss to learn from unlabeled images. It uses two network branches that link semi-supervised classification with semi-supervised segmentation including self-training. The dual-branch approach reduces both the low-level and the high-level artifacts typical when training with few labels. The approach attains significant improvement over existing methods, especially when trained with very few labeled samples. On several standard benchmarks-PASCAL VOC 2012, PASCAL-Context, and Cityscapes-the approach achieves new state-of-the-art in semi-supervised learning.

Structured light range sensors (SLRS) like the Microsoft Kinect have electronic rolling shutters (ERS). The output of such a sensor while in motion is subject to significant motion blur (MB) and rolling shutter (RS) distortion. Most robotic literature still does not explicitly model this distortion, resulting in inaccurate camera motion estimation. In RGBD cameras, we show via experimentation that the distortion undergone by depth images is different from that of color images and provide a mathematical model for it. We propose an algorithm that rectifies for these RS and MB distortions. To assess the performance of the algorithm we conduct an extensive set of experiments for each step of the pipeline. We assess the performance of our algorithm by comparing the performance of the rectified images on scene-flow and camera pose estimation, and show that with our proposed rectification, the performance improvement is significant.

In the field of active perception, object search is a widely studied problem. To search for an object in large rooms, it would be expensive to explore and check each object's similarity with the object of interest. The expense could uncontrollably bloat as the number of objects to be searched increases. If the objects are of the order of a 2-5cm, they appear very small, making it difficult for the present algorithms to recognize them. A general human strategy in such cases is to sparsely identify, from far away (4-6m), if the object of interest is present in the scene. Subsequently, each of the possible objects is analysed from closer proximity to recognize, for further manipulation. In this work, we present a similar framework. We reduce search-space, by identifying existential probability of a small object from a distance followed by a closer 3-D analysis of its point cloud to accurately recognize it. This is achieved by 2-D modelling of the objects using Gaussian Mixture Models followed by recognizing objects using efficient RGB-Depth based algorithm.