Multimodal Large Language Models (MLLMs) have demonstrated impressive performance in short video understanding. However, understanding long-form videos still remains challenging for MLLMs. This paper proposes TimeSuite, a collection of new designs to adapt the existing short-form video MLLMs for long video understanding, including a simple yet efficient framework to process long video sequence, a high-quality video dataset for grounded tuning of MLLMs, and a carefully-designed instruction tuning task to explicitly incorporate the grounding supervision in the traditional QA format. Specifically, based on VideoChat, we propose our long-video MLLM, coined as VideoChat-T, by implementing a token shuffling to compress long video tokens and introducing Temporal Adaptive Position Encoding (TAPE) to enhance the temporal awareness of visual representation. Meanwhile, we introduce the TimePro, a comprehensive grounding-centric instruction tuning dataset composed of 9 tasks and 349k high-quality grounded annotations. Notably, we design a new instruction tuning task type, called Temporal Grounded Caption, to peform detailed video descriptions with the corresponding time stamps prediction. This explicit temporal location prediction will guide MLLM to correctly attend on the visual content when generating description, and thus reduce the hallucination risk caused by the LLMs. Experimental results demonstrate that our TimeSuite provides a successful solution to enhance the long video understanding capability of short-form MLLM, achieving improvement of 5.6% and 6.8% on the benchmarks of Egoschema and VideoMME, respectively. In addition, VideoChat-T exhibits robust zero-shot temporal grounding capabilities, significantly outperforming the existing state-of-the-art MLLMs. After fine-tuning, it performs on par with the traditional supervised expert models.

In today’s software development environment, third-party libraries (TPLs) enhance productivity but also introduce security risks. Effective Software Composition Analysis (SCA) is crucial for managing these risks. Yet, existing SCA tools for C/C++ projects struggle with challenges like detecting modified and nested TPLs, precise version representation, and comprehensive TPL databases. In modern software development, third-party libraries (TPLs) are commonly used to boost functionality and save development time. However, this convenience introduces security risks. We introduce OSSDetector, a new SCA tool that addresses these issues. OSSDetector uses sliding window and fuzzy hashing techniques to generate detailed signatures, improving detection of modified TPLs. It features a "Nested TPL Function Filtering" algorithm to accurately identify and filter nested TPL functions, and a "TPL Recognition" algorithm based on import ratios and function paths to determine the TPLs used in the software. It also addresses version representation by using function weights and release times. To overcome the lack of a comprehensive TPL database, we have developed a large database with 29,416 C/C++ TPLs and 767,405 versions. Experimental results demonstrate that OSSDetector surpasses state-of-the-art tools, achieving better precision (85.52%), recall (79.82%), and F1 score (82.57%), and higher precision (84.27%) at the library version level.

Modern recommender systems employ various sequential modules such as self-attention to learn dynamic user interests. However, these methods are less effective in capturing collaborative and transitional signals within user interaction sequences. First, the self-attention architecture uses the embedding of a single item as the attention query, making it challenging to capture collaborative signals. Second, these methods typically follow an auto-regressive framework, which is unable to learn global item transition patterns. To overcome these limitations, we propose a new method called Multi-Query Self-Attention with Transition-Aware Embedding Distillation (MQSA-TED). First, we propose an L-query self-attention module that employs flexible window sizes for attention queries to capture collaborative signals. In addition, we introduce a multi-query self-attention method that balances the bias-variance trade-off in modeling user preferences by combining long and short-query self-attentions. Second, we develop a transition-aware embedding distillation module that distills global item-to-item transition patterns into item embeddings, which enables the model to memorize and leverage transitional signals and serves as a calibrator for collaborative signals. Experimental results on four real-world datasets demonstrate the effectiveness of the proposed modules.