Phare: A Safety Probe for Large Language Models

Ensuring the safety of large language models (LLMs) is critical for responsible deployment, yet existing evaluations often prioritize performance over identifying failure modes. We introduce Phare, a multilingual diagnostic framework to probe and evaluate LLM behavior across three critical dimensions: hallucination and reliability, social biases, and harmful content generation. Our evaluation of 17 state-of-the-art LLMs reveals patterns of systematic vulnerabilities across all safety dimensions, including sycophancy, prompt sensitivity, and stereotype reproduction. By highlighting these specific failure modes rather than simply ranking models, Phare provides researchers and practitioners with actionable insights to build more robust, aligned, and trustworthy language systems.

RealHarm: A Collection of Real-World Language Model Application Failures

Language model deployments in consumer-facing applications introduce numerous risks. While existing research on harms and hazards of such applications follows top-down approaches derived from regulatory frameworks and theoretical analyses, empirical evidence of real-world failure modes remains underexplored. In this work, we introduce RealHarm, a dataset of annotated problematic interactions with AI agents built from a systematic review of publicly reported incidents. Analyzing harms, causes, and hazards specifically from the deployer's perspective, we find that reputational damage constitutes the predominant organizational harm, while misinformation emerges as the most common hazard category. We empirically evaluate state-of-the-art guardrails and content moderation systems to probe whether such systems would have prevented the incidents, revealing a significant gap in the protection of AI applications.

Object Detection in Aerial Images in Scarce Data Regimes

Most contributions on Few-Shot Object Detection (FSOD) evaluate their methods on natural images only, yet the transferability of the announced performance is not guaranteed for applications on other kinds of images. We demonstrate this with an in-depth analysis of existing FSOD methods on aerial images and observed a large performance gap compared to natural images. Small objects, more numerous in aerial images, are the cause for the apparent performance gap between natural and aerial images. As a consequence, we improve FSOD performance on small objects with a carefully designed attention mechanism. In addition, we also propose a scale-adaptive box similarity criterion, that improves the training and evaluation of FSOD methods, particularly for small objects. We also contribute to generic FSOD with two distinct approaches based on metric learning and fine-tuning. Impressive results are achieved with the fine-tuning method, which encourages tackling more complex scenarios such as Cross-Domain FSOD. We conduct preliminary experiments in this direction and obtain promising results. Finally, we address the deployment of the detection models inside COSE's systems. Detection must be done in real-time in extremely large images (more than 100 megapixels), with limited computation power. Leveraging existing optimization tools such as TensorRT, we successfully tackle this engineering challenge.

Rethinking Intersection Over Union for Small Object Detection in Few-Shot Regime

In Few-Shot Object Detection (FSOD), detecting small objects is extremely difficult. The limited supervision cripples the localization capabilities of the models and a few pixels shift can dramatically reduce the Intersection over Union (IoU) between the ground truth and predicted boxes for small objects. To this end, we propose Scale-adaptive Intersection over Union (SIoU), a novel box similarity measure. SIoU changes with the objects' size, it is more lenient with small object shifts. We conducted a user study and SIoU better aligns than IoU with human judgment. Employing SIoU as an evaluation criterion helps to build more user-oriented models. SIoU can also be used as a loss function to prioritize small objects during training, outperforming existing loss functions. SIoU improves small object detection in the non-few-shot regime, but this setting is unrealistic in the industry as annotated detection datasets are often too expensive to acquire. Hence, our experiments mainly focus on the few-shot regime to demonstrate the superiority and versatility of SIoU loss. SIoU improves significantly FSOD performance on small objects in both natural (Pascal VOC and COCO datasets) and aerial images (DOTA and DIOR). In aerial imagery, small objects are critical and SIoU loss achieves new state-of-the-art FSOD on DOTA and DIOR.

A Comparative Attention Framework for Better Few-Shot Object Detection on Aerial Images

Few-Shot Object Detection (FSOD) methods are mainly designed and evaluated on natural image datasets such as Pascal VOC and MS COCO. However, it is not clear whether the best methods for natural images are also the best for aerial images. Furthermore, direct comparison of performance between FSOD methods is difficult due to the wide variety of detection frameworks and training strategies. Therefore, we propose a benchmarking framework that provides a flexible environment to implement and compare attention-based FSOD methods. The proposed framework focuses on attention mechanisms and is divided into three modules: spatial alignment, global attention, and fusion layer. To remain competitive with existing methods, which often leverage complex training, we propose new augmentation techniques designed for object detection. Using this framework, several FSOD methods are reimplemented and compared. This comparison highlights two distinct performance regimes on aerial and natural images: FSOD performs worse on aerial images. Our experiments suggest that small objects, which are harder to detect in the few-shot setting, account for the poor performance. Finally, we develop a novel multiscale alignment method, Cross-Scales Query-Support Alignment (XQSA) for FSOD, to improve the detection of small objects. XQSA outperforms the state-of-the-art significantly on DOTA and DIOR.

A Unified Framework for Attention-Based Few-Shot Object Detection

Few-Shot Object Detection (FSOD) is a rapidly growing field in computer vision. It consists in finding all occurrences of a given set of classes with only a few annotated examples for each class. Numerous methods have been proposed to address this challenge and most of them are based on attention mechanisms. However, the great variety of classic object detection frameworks and training strategies makes performance comparison between methods difficult. In particular, for attention-based FSOD methods, it is laborious to compare the impact of the different attention mechanisms on performance. This paper aims at filling this shortcoming. To do so, a flexible framework is proposed to allow the implementation of most of the attention techniques available in the literature. To properly introduce such a framework, a detailed review of the existing FSOD methods is firstly provided. Some different attention mechanisms are then reimplemented within the framework and compared with all other parameters fixed.

Experience feedback using Representation Learning for Few-Shot Object Detection on Aerial Images

This paper proposes a few-shot method based on Faster R-CNN and representation learning for object detection in aerial images. The two classification branches of Faster R-CNN are replaced by prototypical networks for online adaptation to new classes. These networks produce embeddings vectors for each generated box, which are then compared with class prototypes. The distance between an embedding and a prototype determines the corresponding classification score. The resulting networks are trained in an episodic manner. A new detection task is randomly sampled at each epoch, consisting in detecting only a subset of the classes annotated in the dataset. This training strategy encourages the network to adapt to new classes as it would at test time. In addition, several ideas are explored to improve the proposed method such as a hard negative examples mining strategy and self-supervised clustering for background objects. The performance of our method is assessed on DOTA, a large-scale remote sensing images dataset. The experiments conducted provide a broader understanding of the capabilities of representation learning. It highlights in particular some intrinsic weaknesses for the few-shot object detection task. Finally, some suggestions and perspectives are formulated according to these insights.