Video Token Merging for Long-form Video Understanding

As the scale of data and models for video understanding rapidly expand, handling long-form video input in transformer-based models presents a practical challenge. Rather than resorting to input sampling or token dropping, which may result in information loss, token merging shows promising results when used in collaboration with transformers. However, the application of token merging for long-form video processing is not trivial. We begin with the premise that token merging should not rely solely on the similarity of video tokens; the saliency of tokens should also be considered. To address this, we explore various video token merging strategies for long-form video classification, starting with a simple extension of image token merging, moving to region-concentrated merging, and finally proposing a learnable video token merging (VTM) algorithm that dynamically merges tokens based on their saliency. Extensive experimental results show that we achieve better or comparable performances on the LVU, COIN, and Breakfast datasets. Moreover, our approach significantly reduces memory costs by 84% and boosts throughput by approximately 6.89 times compared to baseline algorithms.

Automatic programming via large language models with population self-evolution for dynamic job shop scheduling problem

Heuristic dispatching rules (HDRs) are widely regarded as effective methods for solving dynamic job shop scheduling problems (DJSSP) in real-world production environments. However, their performance is highly scenario-dependent, often requiring expert customization. To address this, genetic programming (GP) and gene expression programming (GEP) have been extensively used for automatic algorithm design. Nevertheless, these approaches often face challenges due to high randomness in the search process and limited generalization ability, hindering the application of trained dispatching rules to new scenarios or dynamic environments. Recently, the integration of large language models (LLMs) with evolutionary algorithms has opened new avenues for prompt engineering and automatic algorithm design. To enhance the capabilities of LLMs in automatic HDRs design, this paper proposes a novel population self-evolutionary (SeEvo) method, a general search framework inspired by the self-reflective design strategies of human experts. The SeEvo method accelerates the search process and enhances exploration capabilities. Experimental results show that the proposed SeEvo method outperforms GP, GEP, end-to-end deep reinforcement learning methods, and more than 10 common HDRs from the literature, particularly in unseen and dynamic scenarios.

MambaEviScrib: Mamba and Evidence-Guided Consistency Make CNN Work Robustly for Scribble-Based Weakly Supervised Ultrasound Image Segmentation

Segmenting anatomical structures and lesions from ultrasound images contributes to disease assessment, diagnosis, and treatment. Weakly supervised learning (WSL) based on sparse annotation has achieved encouraging performance and demonstrated the potential to reduce annotation costs. However, ultrasound images often suffer from issues such as poor contrast, unclear edges, as well as varying sizes and locations of lesions. This makes it challenging for convolutional networks with local receptive fields to extract global morphological features from the sparse information provided by scribble annotations. Recently, the visual Mamba based on state space sequence models (SSMs) has significantly reduced computational complexity while ensuring long-range dependencies compared to Transformers. Consequently, for the first time, we apply scribble-based WSL to ultrasound image segmentation and propose a novel hybrid CNN-Mamba framework. Furthermore, due to the characteristics of ultrasound images and insufficient supervision signals, existing consistency regularization often filters out predictions near decision boundaries, leading to unstable predictions of edges. Hence, we introduce the Dempster-Shafer theory (DST) of evidence to devise an Evidence-Guided Consistency (EGC) strategy, which leverages high-evidence predictions more likely to occur near high-density regions to guide low-evidence predictions potentially present near decision boundaries for optimization. During training, the collaboration between the CNN branch and the Mamba branch in the proposed framework draws inspiration from each other based on the EGC strategy. Extensive experiments on four ultrasound public datasets for binary-class and multi-class segmentation demonstrate the competitiveness of the proposed method. The scribble-annotated dataset and code will be made available on https://github.com/GtLinyer/MambaEviScrib.

Benchmarking Sub-Genre Classification For Mainstage Dance Music

Music classification, with a wide range of applications, is one of the most prominent tasks in music information retrieval. To address the absence of comprehensive datasets and high-performing methods in the classification of mainstage dance music, this work introduces a novel benchmark comprising a new dataset and a baseline. Our dataset extends the number of sub-genres to cover most recent mainstage live sets by top DJs worldwide in music festivals. A continuous soft labeling approach is employed to account for tracks that span multiple sub-genres, preserving the inherent sophistication. For the baseline, we developed deep learning models that outperform current state-of-the-art multimodel language models, which struggle to identify house music sub-genres, emphasizing the need for specialized models trained on fine-grained datasets. Our benchmark is applicable to serve for application scenarios such as music recommendation, DJ set curation, and interactive multimedia, where we also provide video demos. Our code is on \url{https://anonymous.4open.science/r/Mainstage-EDM-Benchmark/}.

GANPrompt: Enhancing Robustness in LLM-Based Recommendations with GAN-Enhanced Diversity Prompts

In recent years, LLM has demonstrated remarkable proficiency in comprehending and generating natural language, with a growing prevalence in the domain of recommender systems. However, LLM continues to face a significant challenge in that it is highly susceptible to the influence of prompt words. This inconsistency in response to minor alterations in prompt input may compromise the accuracy and resilience of recommendation models. To address this issue, this paper proposes GANPrompt, a multi-dimensional large language model prompt diversity framework based on Generative Adversarial Networks (GANs). The framework enhances the model's adaptability and stability to diverse prompts by integrating GAN generation techniques with the deep semantic understanding capabilities of LLMs. GANPrompt first trains a generator capable of producing diverse prompts by analysing multidimensional user behavioural data. These diverse prompts are then used to train the LLM to improve its performance in the face of unseen prompts. Furthermore, to ensure a high degree of diversity and relevance of the prompts, this study introduces a mathematical theory-based diversity constraint mechanism that optimises the generated prompts to ensure that they are not only superficially distinct, but also semantically cover a wide range of user intentions. Through extensive experiments on multiple datasets, we demonstrate the effectiveness of the proposed framework, especially in improving the adaptability and robustness of recommender systems in complex and dynamic environments. The experimental results demonstrate that GANPrompt yields substantial enhancements in accuracy and robustness relative to existing state-of-the-art methodologies.

Text-Guided Video Masked Autoencoder

Recent video masked autoencoder (MAE) works have designed improved masking algorithms focused on saliency. These works leverage visual cues such as motion to mask the most salient regions. However, the robustness of such visual cues depends on how often input videos match underlying assumptions. On the other hand, natural language description is an information dense representation of video that implicitly captures saliency without requiring modality-specific assumptions, and has not been explored yet for video MAE. To this end, we introduce a novel text-guided masking algorithm (TGM) that masks the video regions with highest correspondence to paired captions. Without leveraging any explicit visual cues for saliency, our TGM is competitive with state-of-the-art masking algorithms such as motion-guided masking. To further benefit from the semantics of natural language for masked reconstruction, we next introduce a unified framework for joint MAE and masked video-text contrastive learning. We show that across existing masking algorithms, unifying MAE and masked video-text contrastive learning improves downstream performance compared to pure MAE on a variety of video recognition tasks, especially for linear probe. Within this unified framework, our TGM achieves the best relative performance on five action recognition and one egocentric datasets, highlighting the complementary nature of natural language for masked video modeling.

Improving Multi-modal Recommender Systems by Denoising and Aligning Multi-modal Content and User Feedback

Multi-modal recommender systems (MRSs) are pivotal in diverse online web platforms and have garnered considerable attention in recent years. However, previous studies overlook the challenges of (1) noisy multi-modal content, (2) noisy user feedback, and (3) aligning multi-modal content with user feedback. In order to tackle these challenges, we propose Denoising and Aligning Multi-modal Recommender System (DA-MRS). To mitigate multi-modal noise, DA-MRS first constructs item-item graphs determined by consistent content similarity across modalities. To denoise user feedback, DA-MRS associates the probability of observed feedback with multi-modal content and devises a denoised BPR loss. Furthermore, DA-MRS implements Alignment guided by User preference to enhance task-specific item representation and Alignment guided by graded Item relations to provide finer-grained alignment. Extensive experiments verify that DA-MRS is a plug-and-play framework and achieves significant and consistent improvements across various datasets, backbone models, and noisy scenarios.

Unraveling and Mitigating Retriever Inconsistencies in Retrieval-Augmented Large Language Models

Although Retrieval-Augmented Large Language Models (RALMs) demonstrate their superiority in terms of factuality, they do not consistently outperform the original retrieval-free Language Models (LMs). Our experiments reveal that this example-level performance inconsistency exists not only between retrieval-augmented and retrieval-free LM but also among different retrievers. To understand this phenomenon, we investigate the degeneration behavior of RALMs and theoretically decompose it into four categories. Further analysis based on our decomposition reveals that the innate difference in knowledge sources and the unpredictable degeneration of the reader model contribute most to the inconsistency. Drawing from our analysis, we introduce Ensemble of Retrievers (EoR), a trainable framework that can adaptively retrieve from different knowledge sources and effectively decrease unpredictable reader errors. Our experiments on Open Domain Question Answering show that EoR substantially improves performance over the RALM with a single retriever by considerably reducing inconsistent behaviors.

A Deep Learning-Driven Pipeline for Differentiating Hypertrophic Cardiomyopathy from Cardiac Amyloidosis Using 2D Multi-View Echocardiography

Hypertrophic cardiomyopathy (HCM) and cardiac amyloidosis (CA) are both heart conditions that can progress to heart failure if untreated. They exhibit similar echocardiographic characteristics, often leading to diagnostic challenges. This paper introduces a novel multi-view deep learning approach that utilizes 2D echocardiography for differentiating between HCM and CA. The method begins by classifying 2D echocardiography data into five distinct echocardiographic views: apical 4-chamber, parasternal long axis of left ventricle, parasternal short axis at levels of the mitral valve, papillary muscle, and apex. It then extracts features of each view separately and combines five features for disease classification. A total of 212 patients diagnosed with HCM, and 30 patients diagnosed with CA, along with 200 individuals with normal cardiac function(Normal), were enrolled in this study from 2018 to 2022. This approach achieved a precision, recall of 0.905, and micro-F1 score of 0.904, demonstrating its effectiveness in accurately identifying HCM and CA using a multi-view analysis.

Personalized Language Modeling from Personalized Human Feedback

Reinforcement Learning from Human Feedback (RLHF) is the current dominating framework to fine-tune large language models to better align with human preferences. However, the underlying premise of algorithms developed under this framework can be problematic when user preferences encoded in human feedback are diverse. In this work, we aim to address this problem by developing methods for building personalized language models. We first formally introduce the task of learning from personalized human feedback and explain why vanilla RLHF can be problematic in this context. We then propose a general Personalized-RLHF (P-RLHF) framework, which requires one to jointly learn a user model and a language (or reward) model. The user model takes in user information and outputs user representations. Its structure encodes our assumptions about user preferences underlying the feedback data. We develop new learning objectives for personalized reward modeling and personalized Direct Preference Optimization. To demonstrate the efficacy of our method, we test it on real-world text summarization data with annotated preferences and annotator information. We fine-tune GPT-J 6B to obtain personalized language (and reward) models, which outperform non-personalized models in terms of aligning with individual preferences.