MERLIN: Multimodal Embedding Refinement via LLM-based Iterative Navigation for Text-Video Retrieval-Rerank Pipeline

The rapid expansion of multimedia content has made accurately retrieving relevant videos from large collections increasingly challenging. Recent advancements in text-video retrieval have focused on cross-modal interactions, large-scale foundation model training, and probabilistic modeling, yet often neglect the crucial user perspective, leading to discrepancies between user queries and the content retrieved. To address this, we introduce MERLIN (Multimodal Embedding Refinement via LLM-based Iterative Navigation), a novel, training-free pipeline that leverages Large Language Models (LLMs) for iterative feedback learning. MERLIN refines query embeddings from a user perspective, enhancing alignment between queries and video content through a dynamic question answering process. Experimental results on datasets like MSR-VTT, MSVD, and ActivityNet demonstrate that MERLIN substantially improves Recall@1, outperforming existing systems and confirming the benefits of integrating LLMs into multimodal retrieval systems for more responsive and context-aware multimedia retrieval.

Practical Dataset Distillation Based on Deep Support Vectors

Conventional dataset distillation requires significant computational resources and assumes access to the entire dataset, an assumption impractical as it presumes all data resides on a central server. In this paper, we focus on dataset distillation in practical scenarios with access to only a fraction of the entire dataset. We introduce a novel distillation method that augments the conventional process by incorporating general model knowledge via the addition of Deep KKT (DKKT) loss. In practical settings, our approach showed improved performance compared to the baseline distribution matching distillation method on the CIFAR-10 dataset. Additionally, we present experimental evidence that Deep Support Vectors (DSVs) offer unique information to the original distillation, and their integration results in enhanced performance.

Do not think pink elephant!

Large Models (LMs) have heightened expectations for the potential of general AI as they are akin to human intelligence. This paper shows that recent large models such as Stable Diffusion and DALL-E3 also share the vulnerability of human intelligence, namely the "white bear phenomenon". We investigate the causes of the white bear phenomenon by analyzing their representation space. Based on this analysis, we propose a simple prompt-based attack method, which generates figures prohibited by the LM provider's policy. To counter these attacks, we introduce prompt-based defense strategies inspired by cognitive therapy techniques, successfully mitigating attacks by up to 48.22\%.

Coreset Selection for Object Detection

Coreset selection is a method for selecting a small, representative subset of an entire dataset. It has been primarily researched in image classification, assuming there is only one object per image. However, coreset selection for object detection is more challenging as an image can contain multiple objects. As a result, much research has yet to be done on this topic. Therefore, we introduce a new approach, Coreset Selection for Object Detection (CSOD). CSOD generates imagewise and classwise representative feature vectors for multiple objects of the same class within each image. Subsequently, we adopt submodular optimization for considering both representativeness and diversity and utilize the representative vectors in the submodular optimization process to select a subset. When we evaluated CSOD on the Pascal VOC dataset, CSOD outperformed random selection by +6.4%p in AP$_{50}$ when selecting 200 images.

Semi-Supervised Domain Adaptation for Wildfire Detection

Recently, both the frequency and intensity of wildfires have increased worldwide, primarily due to climate change. In this paper, we propose a novel protocol for wildfire detection, leveraging semi-supervised Domain Adaptation for object detection, accompanied by a corresponding dataset designed for use by both academics and industries. Our dataset encompasses 30 times more diverse labeled scenes for the current largest benchmark wildfire dataset, HPWREN, and introduces a new labeling policy for wildfire detection. Inspired by CoordConv, we propose a robust baseline, Location-Aware Object Detection for Semi-Supervised Domain Adaptation (LADA), utilizing a teacher-student based framework capable of extracting translational variance features characteristic of wildfires. With only using 1% target domain labeled data, our framework significantly outperforms our source-only baseline by a notable margin of 3.8% in mean Average Precision on the HPWREN wildfire dataset. Our dataset is available at https://github.com/BloomBerry/LADA.

Unleash the Potential of CLIP for Video Highlight Detection

Multimodal and large language models (LLMs) have revolutionized the utilization of open-world knowledge, unlocking novel potentials across various tasks and applications. Among these domains, the video domain has notably benefited from their capabilities. In this paper, we present Highlight-CLIP (HL-CLIP), a method designed to excel in the video highlight detection task by leveraging the pre-trained knowledge embedded in multimodal models. By simply fine-tuning the multimodal encoder in combination with our innovative saliency pooling technique, we have achieved the state-of-the-art performance in the highlight detection task, the QVHighlight Benchmark, to the best of our knowledge.

Deep Support Vectors

While the success of deep learning is commonly attributed to its theoretical equivalence with Support Vector Machines (SVM), the practical implications of this relationship have not been thoroughly explored. This paper pioneers an exploration in this domain, specifically focusing on the identification of Deep Support Vectors (DSVs) within deep learning models. We introduce the concept of DeepKKT conditions, an adaptation of the traditional Karush-Kuhn-Tucker (KKT) conditions tailored for deep learning. Through empirical investigations, we illustrate that DSVs exhibit similarities to support vectors in SVM, offering a tangible method to interpret the decision-making criteria of models. Additionally, our findings demonstrate that models can be effectively reconstructed using DSVs, resembling the process in SVM. The code will be available.

Harmonizing Visual and Textual Embeddings for Zero-Shot Text-to-Image Customization

In a surge of text-to-image (T2I) models and their customization methods that generate new images of a user-provided subject, current works focus on alleviating the costs incurred by a lengthy per-subject optimization. These zero-shot customization methods encode the image of a specified subject into a visual embedding which is then utilized alongside the textual embedding for diffusion guidance. The visual embedding incorporates intrinsic information about the subject, while the textual embedding provides a new, transient context. However, the existing methods often 1) are significantly affected by the input images, eg., generating images with the same pose, and 2) exhibit deterioration in the subject's identity. We first pin down the problem and show that redundant pose information in the visual embedding interferes with the textual embedding containing the desired pose information. To address this issue, we propose orthogonal visual embedding which effectively harmonizes with the given textual embedding. We also adopt the visual-only embedding and inject the subject's clear features utilizing a self-attention swap. Our results demonstrate the effectiveness and robustness of our method, which offers highly flexible zero-shot generation while effectively maintaining the subject's identity.

HourglassNeRF: Casting an Hourglass as a Bundle of Rays for Few-shot Neural Rendering

Recent advancements in the Neural Radiance Field (NeRF) have bolstered its capabilities for novel view synthesis, yet its reliance on dense multi-view training images poses a practical challenge. Addressing this, we propose HourglassNeRF, an effective regularization-based approach with a novel hourglass casting strategy. Our proposed hourglass is conceptualized as a bundle of additional rays within the area between the original input ray and its corresponding reflection ray, by featurizing the conical frustum via Integrated Positional Encoding (IPE). This design expands the coverage of unseen views and enables an adaptive high-frequency regularization based on target pixel photo-consistency. Furthermore, we propose luminance consistency regularization based on the Lambertian assumption, which is known to be effective for training a set of augmented rays under the few-shot setting. Leveraging the inherent property of a Lambertian surface, which retains consistent luminance irrespective of the viewing angle, we assume our proposed hourglass as a collection of flipped diffuse reflection rays and enhance the luminance consistency between the original input ray and its corresponding hourglass, resulting in more physically grounded training framework and performance improvement. Our HourglassNeRF outperforms its baseline and achieves competitive results on multiple benchmarks with sharply rendered fine details. The code will be available.

Mitigating the Bias in the Model for Continual Test-Time Adaptation

Continual Test-Time Adaptation (CTA) is a challenging task that aims to adapt a source pre-trained model to continually changing target domains. In the CTA setting, a model does not know when the target domain changes, thus facing a drastic change in the distribution of streaming inputs during the test-time. The key challenge is to keep adapting the model to the continually changing target domains in an online manner. We find that a model shows highly biased predictions as it constantly adapts to the chaining distribution of the target data. It predicts certain classes more often than other classes, making inaccurate over-confident predictions. This paper mitigates this issue to improve performance in the CTA scenario. To alleviate the bias issue, we make class-wise exponential moving average target prototypes with reliable target samples and exploit them to cluster the target features class-wisely. Moreover, we aim to align the target distributions to the source distribution by anchoring the target feature to its corresponding source prototype. With extensive experiments, our proposed method achieves noteworthy performance gain when applied on top of existing CTA methods without substantial adaptation time overhead.