Abstract:While Multimodal Large Language Models (MLLMs) demonstrate remarkable capabilities on static images, they often fall short in comprehending dynamic, information-dense short-form videos, a dominant medium in today's digital landscape. To bridge this gap, we introduce \textbf{Kwai Keye-VL}, an 8-billion-parameter multimodal foundation model engineered for leading-edge performance in short-video understanding while maintaining robust general-purpose vision-language abilities. The development of Keye-VL rests on two core pillars: a massive, high-quality dataset exceeding 600 billion tokens with a strong emphasis on video, and an innovative training recipe. This recipe features a four-stage pre-training process for solid vision-language alignment, followed by a meticulous two-phase post-training process. The first post-training stage enhances foundational capabilities like instruction following, while the second phase focuses on stimulating advanced reasoning. In this second phase, a key innovation is our five-mode ``cold-start'' data mixture, which includes ``thinking'', ``non-thinking'', ``auto-think'', ``think with image'', and high-quality video data. This mixture teaches the model to decide when and how to reason. Subsequent reinforcement learning (RL) and alignment steps further enhance these reasoning capabilities and correct abnormal model behaviors, such as repetitive outputs. To validate our approach, we conduct extensive evaluations, showing that Keye-VL achieves state-of-the-art results on public video benchmarks and remains highly competitive on general image-based tasks (Figure 1). Furthermore, we develop and release the \textbf{KC-MMBench}, a new benchmark tailored for real-world short-video scenarios, where Keye-VL shows a significant advantage.
Abstract:The advent of autonomous agents is transforming interactions with Graphical User Interfaces (GUIs) by employing natural language as a powerful intermediary. Despite the predominance of Supervised Fine-Tuning (SFT) methods in current GUI agents for achieving spatial localization, these methods face substantial challenges due to their limited capacity to accurately perceive positional data. Existing strategies, such as reinforcement learning, often fail to assess positional accuracy effectively, thereby restricting their utility. In response, we introduce Location Preference Optimization (LPO), a novel approach that leverages locational data to optimize interaction preferences. LPO uses information entropy to predict interaction positions by focusing on zones rich in information. Besides, it further introduces a dynamic location reward function based on physical distance, reflecting the varying importance of interaction positions. Supported by Group Relative Preference Optimization (GRPO), LPO facilitates an extensive exploration of GUI environments and significantly enhances interaction precision. Comprehensive experiments demonstrate LPO's superior performance, achieving SOTA results across both offline benchmarks and real-world online evaluations. Our code will be made publicly available soon, at https://github.com/AIDC-AI/LPO.
Abstract:In recent years, integrated short-video and live-streaming platforms have gained massive global adoption, offering dynamic content creation and consumption. Unlike pre-recorded short videos, live-streaming enables real-time interaction between authors and users, fostering deeper engagement. However, this dynamic nature introduces a critical challenge for recommendation systems (RecSys): the same live-streaming vastly different experiences depending on when a user watching. To optimize recommendations, a RecSys must accurately interpret the real-time semantics of live content and align them with user preferences.
Abstract:Mainstream test-time adaptation (TTA) techniques endeavor to mitigate distribution shifts via entropy minimization for multi-class classification, inherently increasing the probability of the most confident class. However, when encountering multi-label instances, the primary challenge stems from the varying number of labels per image, and prioritizing only the highest probability class inevitably undermines the adaptation of other positive labels. To address this issue, we investigate TTA within multi-label scenario (ML--TTA), developing Bound Entropy Minimization (BEM) objective to simultaneously increase the confidence of multiple top predicted labels. Specifically, to determine the number of labels for each augmented view, we retrieve a paired caption with yielded textual labels for that view. These labels are allocated to both the view and caption, called weak label set and strong label set with the same size k. Following this, the proposed BEM considers the highest top-k predicted labels from view and caption as a single entity, respectively, learning both view and caption prompts concurrently. By binding top-k predicted labels, BEM overcomes the limitation of vanilla entropy minimization, which exclusively optimizes the most confident class. Across the MSCOCO, VOC, and NUSWIDE multi-label datasets, our ML--TTA framework equipped with BEM exhibits superior performance compared to the latest SOTA methods, across various model architectures, prompt initialization, and varying label scenarios. The code is available at https://github.com/Jinx630/ML-TTA.
Abstract:Treating texts as images, combining prompts with textual labels for prompt tuning, and leveraging the alignment properties of CLIP have been successfully applied in zero-shot multi-label image recognition. Nonetheless, relying solely on textual labels to store visual information is insufficient for representing the diversity of visual objects. In this paper, we propose reversing the training process of CLIP and introducing the concept of Pseudo Visual Prompts. These prompts are initialized for each object category and pre-trained on large-scale, low-cost sentence data generated by large language models. This process mines the aligned visual information in CLIP and stores it in class-specific visual prompts. We then employ contrastive learning to transfer the stored visual information to the textual labels, enhancing their visual representation capacity. Additionally, we introduce a dual-adapter module that simultaneously leverages knowledge from the original CLIP and new learning knowledge derived from downstream datasets. Benefiting from the pseudo visual prompts, our method surpasses the state-of-the-art not only on clean annotated text data but also on pseudo text data generated by large language models.
Abstract:In this paper, we present our solution for the WSDM2023 Toloka Visual Question Answering Challenge. Inspired by the application of multimodal pre-trained models to various downstream tasks(e.g., visual question answering, visual grounding, and cross-modal retrieval), we approached this competition as a visual grounding task, where the input is an image and a question, guiding the model to answer the question and display the answer as a bounding box on the image. We designed a three-stage solution for this task. Specifically, we used the visual-language pre-trained model OFA as the foundation. In the first stage, we constructed a large-scale synthetic dataset similar to the competition dataset and coarse-tuned the model to learn generalized semantic information. In the second stage, we treated the competition task as a visual grounding task, loaded the weights from the previous stage, and continued to fine-tune the model on the competition dataset, transferring the semantic information learned in the first stage to the competition task. Finally, we designed a bounding box matching and replacing post-processing strategy to correct the model's prediction results. Our team achieved a score of 76.342 on the final leaderboard, ranking second.
Abstract:This report introduces a solution to The task of RGB-TIR object detection from the perspective of unmanned aerial vehicles. Unlike traditional object detection methods, RGB-TIR object detection aims to utilize both RGB and TIR images for complementary information during detection. The challenges of RGB-TIR object detection from the perspective of unmanned aerial vehicles include highly complex image backgrounds, frequent changes in lighting, and uncalibrated RGB-TIR image pairs. To address these challenges at the model level, we utilized a lightweight YOLOv9 model with extended multi-level auxiliary branches that enhance the model's robustness, making it more suitable for practical applications in unmanned aerial vehicle scenarios. For image fusion in RGB-TIR detection, we incorporated a fusion module into the backbone network to fuse images at the feature level, implicitly addressing calibration issues. Our proposed method achieved an mAP score of 0.516 and 0.543 on A and B benchmarks respectively while maintaining the highest inference speed among all models.
Abstract:In this paper, we present our champion solution to the Global Artificial Intelligence Technology Innovation Competition Track 1: Medical Imaging Diagnosis Report Generation. We select CPT-BASE as our base model for the text generation task. During the pre-training stage, we delete the mask language modeling task of CPT-BASE and instead reconstruct the vocabulary, adopting a span mask strategy and gradually increasing the number of masking ratios to perform the denoising auto-encoder pre-training task. In the fine-tuning stage, we design iterative retrieval augmentation and noise-aware similarity bucket prompt strategies. The retrieval augmentation constructs a mini-knowledge base, enriching the input information of the model, while the similarity bucket further perceives the noise information within the mini-knowledge base, guiding the model to generate higher-quality diagnostic reports based on the similarity prompts. Surprisingly, our single model has achieved a score of 2.321 on leaderboard A, and the multiple model fusion scores are 2.362 and 2.320 on the A and B leaderboards respectively, securing first place in the rankings.
Abstract:In this paper, we propose a solution for improving the quality of temporal sound localization. We employ a multimodal fusion approach to combine visual and audio features. High-quality visual features are extracted using a state-of-the-art self-supervised pre-training network, resulting in efficient video feature representations. At the same time, audio features serve as complementary information to help the model better localize the start and end of sounds. The fused features are trained in a multi-scale Transformer for training. In the final test dataset, we achieved a mean average precision (mAP) of 0.33, obtaining the second-best performance in this track.
Abstract:The recent introduction of prompt tuning based on pre-trained vision-language models has dramatically improved the performance of multi-label image classification. However, some existing strategies that have been explored still have drawbacks, i.e., either exploiting massive labeled visual data at a high cost or using text data only for text prompt tuning and thus failing to learn the diversity of visual knowledge. Hence, the application scenarios of these methods are limited. In this paper, we propose a pseudo-visual prompt~(PVP) module for implicit visual prompt tuning to address this problem. Specifically, we first learn the pseudo-visual prompt for each category, mining diverse visual knowledge by the well-aligned space of pre-trained vision-language models. Then, a co-learning strategy with a dual-adapter module is designed to transfer visual knowledge from pseudo-visual prompt to text prompt, enhancing their visual representation abilities. Experimental results on VOC2007, MS-COCO, and NUSWIDE datasets demonstrate that our method can surpass state-of-the-art~(SOTA) methods across various settings for multi-label image classification tasks. The code is available at https://github.com/njustkmg/PVP.