Search

Recent advancements in video autoencoders (Video AEs) have significantly improved the quality and efficiency of video generation. In this paper, we propose a novel and compact video autoencoder, VidTwin, that decouples video into two distinct latent spaces: Structure latent vectors, which capture overall content and global movement, and Dynamics latent vectors, which represent fine-grained details and rapid movements. Specifically, our approach leverages an Encoder-Decoder backbone, augmented with two submodules for extracting these latent spaces, respectively. The first submodule employs a Q-Former to extract low-frequency motion trends, followed by downsampling blocks to remove redundant content details. The second averages the latent vectors along the spatial dimension to capture rapid motion. Extensive experiments show that VidTwin achieves a high compression rate of 0.20% with high reconstruction quality (PSNR of 28.14 on the MCL-JCV dataset), and performs efficiently and effectively in downstream generative tasks. Moreover, our model demonstrates explainability and scalability, paving the way for future research in video latent representation and generation. Our code has been released at https://github.com/microsoft/VidTok/tree/main/vidtwin.

Multimodal punchlines, which involve humor or sarcasm conveyed in image-caption pairs, are a popular way of communication on online multimedia platforms. With the rapid development of multimodal large language models (MLLMs), it is essential to assess their ability to effectively comprehend these punchlines. However, existing benchmarks on punchline comprehension suffer from three major limitations: 1) language shortcuts that allow models to solely rely on text, 2) lack of question diversity, and 3) narrow focus on a specific domain of multimodal content (e.g., cartoon). To address these limitations, we introduce a multimodal \textbf{Punch}line comprehension \textbf{Bench}mark, named \textbf{PunchBench}, which is tailored for accurate and comprehensive evaluation of punchline comprehension. To enhance the evaluation accuracy, we generate synonymous and antonymous captions by modifying original captions, which mitigates the impact of shortcuts in the captions. To provide a comprehensive evaluation, PunchBench incorporates diverse question formats and image-captions from various domains. On this basis, we conduct extensive evaluations and reveal a significant gap between state-of-the-art MLLMs and humans in punchline comprehension. To improve punchline comprehension, we propose Simple-to-Complex Chain-of-Question (SC-CoQ) strategy, enabling the models to incrementally address complicated questions by first mastering simple ones. SC-CoQ effectively enhances the performance of various MLLMs on PunchBench, surpassing in-context learning and chain-of-thought.

With consecutive bands in a wide range of wavelengths, hyperspectral images (HSI) have provided a unique tool for object detection task. However, existing HSI object detection methods have not been fully utilized in real applications, which is mainly resulted by the difference of spatial and spectral resolution between the unlabeled target domain and a labeled source domain, i.e. the domain shift of HSI. In this work, we aim to explore the unsupervised cross-domain object detection of HSI. Our key observation is that the local spatial-spectral characteristics remain invariant across different domains. For solving the problem of domain-shift, we propose a HSI cross-domain object detection method based on spectral-spatial feature alignment, which is the first attempt in the object detection community to the best of our knowledge. Firstly, we develop a spectral-spatial alignment module to extract domain-invariant local spatial-spectral features. Secondly, the spectral autocorrelation module has been designed to solve the domain shift in the spectral domain specifically, which can effectively align HSIs with different spectral resolutions. Besides, we have collected and annotated an HSI dataset for the cross-domain object detection. Our experimental results have proved the effectiveness of HSI cross-domain object detection, which has firstly demonstrated a significant and promising step towards HSI cross-domain object detection in the object detection community.

The proximal bundle algorithm (PBA) is a fundamental and computationally effective algorithm for solving optimization problems with nonsmooth components. We investigate its convergence rate, focusing on composite settings where one function is smooth and the other is piecewise linear. We interpret a sequence of null steps of the PBA as a Frank-Wolfe algorithm on the Moreau envelope of the dual problem. In light of this correspondence, we first extend the linear convergence of Kelley's method on convex piecewise linear functions from the positive homogeneous to the general case. Building on this result, we propose a novel complexity analysis of PBA and derive a $\mathcal{O}(\epsilon^{-4/5})$ iteration complexity, improving upon the best known $\mathcal{O}(\epsilon^{-2})$ guarantee. This approach also unveils new insights on bundle management.

Explainable molecular property prediction is essential for various scientific fields, such as drug discovery and material science. Despite delivering intrinsic explainability, linear models struggle with capturing complex, non-linear patterns. Large language models (LLMs), on the other hand, yield accurate predictions through powerful inference capabilities yet fail to provide chemically meaningful explanations for their predictions. This work proposes a novel framework, called MoleX, which leverages LLM knowledge to build a simple yet powerful linear model for accurate molecular property prediction with faithful explanations. The core of MoleX is to model complicated molecular structure-property relationships using a simple linear model, augmented by LLM knowledge and a crafted calibration strategy. Specifically, to extract the maximum amount of task-relevant knowledge from LLM embeddings, we employ information bottleneck-inspired fine-tuning and sparsity-inducing dimensionality reduction. These informative embeddings are then used to fit a linear model for explainable inference. Moreover, we introduce residual calibration to address prediction errors stemming from linear models' insufficient expressiveness of complex LLM embeddings, thus recovering the LLM's predictive power and boosting overall accuracy. Theoretically, we provide a mathematical foundation to justify MoleX's explainability. Extensive experiments demonstrate that MoleX outperforms existing methods in molecular property prediction, establishing a new milestone in predictive performance, explainability, and efficiency. In particular, MoleX enables CPU inference and accelerates large-scale dataset processing, achieving comparable performance 300x faster with 100,000 fewer parameters than LLMs. Additionally, the calibration improves model performance by up to 12.7% without compromising explainability.

Video Large Language Models (Video LLMs) have shown promising capabilities in video comprehension, yet they struggle with tracking temporal changes and reasoning about temporal relationships. While previous research attributed this limitation to the ineffective temporal encoding of visual inputs, our diagnostic study reveals that video representations contain sufficient information for even small probing classifiers to achieve perfect accuracy. Surprisingly, we find that the key bottleneck in Video LLMs' temporal reasoning capability stems from the underlying LLM's inherent difficulty with temporal concepts, as evidenced by poor performance on textual temporal question-answering tasks. Building on this discovery, we introduce the Textual Temporal reasoning Transfer (T3). T3 synthesizes diverse temporal reasoning tasks in pure text format from existing image-text datasets, addressing the scarcity of video samples with complex temporal scenarios. Remarkably, without using any video data, T3 enhances LongVA-7B's temporal understanding, yielding a 5.3 absolute accuracy improvement on the challenging TempCompass benchmark, which enables our model to outperform ShareGPT4Video-8B trained on 28,000 video samples. Additionally, the enhanced LongVA-7B model achieves competitive performance on comprehensive video benchmarks. For example, it achieves a 49.7 accuracy on the Temporal Reasoning task of Video-MME, surpassing powerful large-scale models such as InternVL-Chat-V1.5-20B and VILA1.5-40B. Further analysis reveals a strong correlation between textual and video temporal task performance, validating the efficacy of transferring temporal reasoning abilities from text to video domains., Comment: Project page: https://video-t3.github.io

In the financial risk domain, particularly in credit default prediction and fraud detection, accurate identification of high-risk class instances is paramount, as their occurrence can have significant economic implications. Although machine learning models have gained widespread adoption for risk prediction, their performance is often hindered by the scarcity and diversity of high-quality data. This limitation stems from factors in datasets such as small risk sample sizes, high labeling costs, and severe class imbalance, which impede the models' ability to learn effectively and accurately forecast critical events. This study investigates data pre-processing techniques to enhance existing financial risk datasets by introducing TriEnhance, a straightforward technique that entails: (1) generating synthetic samples specifically tailored to the minority class, (2) filtering using binary feedback to refine samples, and (3) self-learning with pseudo-labels. Our experiments across six benchmark datasets reveal the efficacy of TriEnhance, with a notable focus on improving minority class calibration, a key factor for developing more robust financial risk prediction systems.

For Mixture-of-Experts (MoE) models, an unbalanced expert load will lead to routing collapse or increased computational overhead. Existing methods commonly employ an auxiliary loss to encourage load balance, but a large auxiliary loss will introduce non-negligible interference gradients into training and thus impair the model performance. In order to control load balance while not producing undesired gradients during training, we propose Loss-Free Balancing, featured by an auxiliary-loss-free load balancing strategy. To be specific, before the top-K routing decision, Loss-Free Balancing will first apply an expert-wise bias to the routing scores of each expert. By dynamically updating the bias of each expert according to its recent load, Loss-Free Balancing can consistently maintain a balanced distribution of expert load. In addition, since Loss-Free Balancing does not produce any interference gradients, it also elevates the upper bound of model performance gained from MoE training. We validate the performance of Loss-Free Balancing on MoE models with up to 3B parameters trained on up to 200B tokens. Experimental results show that Loss-Free Balancing achieves both better performance and better load balance compared with traditional auxiliary-loss-controlled load balancing strategies.

In the paper, we discuss the development of the multi-gap resistive plate chamber Time-of-Flight (TOF) technology and the production of the STAR TOF detector in China at the beginning of the 21st century. Then we review recent experimental results from the first beam energy scan program (BES-I) at the Relativistic Heavy Ion Collider (RHIC). Topics cover measurements of collectivity, chirality, criticality, global polarization, strangeness, heavy-flavor, di-lepton and light nuclei productions., Comment: 31 pages, 33 figures. This review is dedicated to Professor Wenqing Shen on the occasion to celebrate his leadership of the Chinese STAR Collaboration, the development and production of the STAR MRPC TOF detector in China and many physics analyses

As an approximate Goldstone boson with zero quantum number and zero standard model charge, the decay processes of long-lived $\eta$ meson offer a unique opportunity to explore new physics beyond the standard model and new sources of CP violation, as well as test the low-energy QCD theory and measure the fundamental parameters of light quarks. To pursue these goals in the physics frontiers, we propose a plan to construct a super $\eta$ factory at HIAF high-energy terminal or at CiADS after its energy upgrade. The high-intensity proton beam at HIAF enables the production of a vast number of $\eta$ samples, exceeding $10^{13}$ events per year in the first stage, utilizing multiple layers of thin targets made of light nucleus. This paper presents the physics goals, the first-version conceptual design of the spectrometer, and some preliminary simulation results., Comment: 23 pages, 13 figures

The electric vehicle (EV) battery supply chain will face challenges in sourcing scarce, expensive minerals required for manufacturing and in disposing of hazardous retired batteries. Integrating recycling technology into the supply chain has the potential to alleviate these issues; however, players in the battery market must design investment plans for recycling facilities. In this paper, we propose a two-stage stochastic optimization model for computing minimum cost recycling capacity decisions, in which retired batteries are recycled and recovered materials are used to manufacture new batteries. The model is a separable concave minimization subject to linear constraints, a class for which we design a new finitely convergent global optimization algorithm based on piecewise linear approximation that solves up to 10x faster than comparable algorithms. We propose an equivalent reformulation of the model that reduces the total number of variables by introducing integrality constraints. The reformulation can also be solved by our global algorithm with drastically reduced solve times. We detail a cut grouping strategy for Benders' decomposition in the second stage which improves convergence relative to single-cut and multi-cut implementations. To produce a set of second-stage scenarios, we design an approach for generating time-series projections for new battery demand, retired battery supply, and material costs. Analysis of the optimal solutions shows that effective investment in recycling can reduce battery manufacturing costs by 22% and reduce environmental impacts by up to 7%.

The visual projector, which bridges the vision and language modalities and facilitates cross-modal alignment, serves as a crucial component in MLLMs. However, measuring the effectiveness of projectors in vision-language alignment remains under-explored, which currently can only be inferred from the performance of MLLMs on downstream tasks. Motivated by the problem, this study examines the projector module by interpreting the vision-language semantic flow within MLLMs. Specifically, we trace back the semantic relevance flow from generated language tokens to raw visual encoder patches and the intermediate outputs produced by projectors. Our findings reveal that compressive projectors (e.g., QFormer), abstract visual patches into a limited set of semantic concepts, such as objects or attributes, resulting in a 'double abstraction' phenomenon. This involves a first visual semantic abstraction by the projector referring to pre-defined query tokens, and a second extraction by the LLM based on text instructions. The double abstraction is inefficient in training and will result in cumulative vision semantics deficiency. To mitigate this issue, we propose the key insight of 'Decouple Compression from Abstraction (DeCo), that is compressing the visual token number at the patch level by projectors and allowing the LLM to handle visual semantic abstraction entirely. Consequently, we adopt a simple compressor, i.e., 2D Adaptive Pooling, to downsample visual patches in a parameter-free manner. Empirical evaluation demonstrates that DeCo surpasses traditional compressive projectors regarding both performance and efficiency. It achieves performance gains of 0.9%, 7.1%, and 2.9% across the MLLM Benchmarks, Visual Localization, and Open-ended VQA tasks with fewer trainable parameters and faster convergence speed.

Recent talking avatar generation models have made strides in achieving realistic and accurate lip synchronization with the audio, but often fall short in controlling and conveying detailed expressions and emotions of the avatar, making the generated video less vivid and controllable. In this paper, we propose a novel text-guided approach for generating emotionally expressive 2D avatars, offering fine-grained control, improved interactivity, and generalizability to the resulting video. Our framework, named InstructAvatar, leverages a natural language interface to control the emotion as well as the facial motion of avatars. Technically, we design an automatic annotation pipeline to construct an instruction-video paired training dataset, equipped with a novel two-branch diffusion-based generator to predict avatars with audio and text instructions at the same time. Experimental results demonstrate that InstructAvatar produces results that align well with both conditions, and outperforms existing methods in fine-grained emotion control, lip-sync quality, and naturalness. Our project page is https://wangyuchi369.github.io/InstructAvatar/., Comment: Project page: https://wangyuchi369.github.io/InstructAvatar/

Coherent supercontinuum radiation in the extreme-ultraviolet (XUV) range is indispensable for synthesizing attosecond light pulses and for exploring transient atomic structures. Here, we report the striking observations of coherent XUV supercontinuum (XSC) extended from below to far above the ionization threshold, which exhibits completely different temporal and spatial properties comparing to the conventional rescattering induced high harmonic generation (HHG). We demonstrate that the strong-field created coherence among bound orbitals strongly distort the atomic transition energies during the pulse, leading to coherent emission spanning tens of electron-volts, in contrast to the line emission via free-induction decay occurring after the pulse. The supposed non-radiating bound dark states contribute as well by emitting dressed energy through dark-to-bright emission mechanism. All the processes modulated at sub-cycle time scale jointly form this new-type coherent XSC. This work achieves the strong-field attosecond control of the exotic atomic radiation dynamics and provides the means of simultaneous generation of separated attosecond sources, i.e., XSC and HHG, with potential advancing attosecond interferometry.

Diffusion models have exhibited remarkable capabilities in text-to-image generation. However, their performance in image-to-text generation, specifically image captioning, has lagged behind Auto-Regressive (AR) models, casting doubt on their applicability for such tasks. In this work, we revisit diffusion models, highlighting their capacity for holistic context modeling and parallel decoding. With these benefits, diffusion models can alleviate the inherent limitations of AR methods, including their slow inference speed, error propagation, and unidirectional constraints. Furthermore, we identify the prior underperformance of diffusion models stemming from the absence of an effective latent space for image-text alignment, and the discrepancy between continuous diffusion processes and discrete textual data. In response, we introduce a novel architecture, LaDiC, which utilizes a split BERT to create a dedicated latent space for captions and integrates a regularization module to manage varying text lengths. Our framework also includes a diffuser for semantic image-to-text conversion and a Back&Refine technique to enhance token interactivity during inference. LaDiC achieves state-of-the-art performance for diffusion-based methods on the MS COCO dataset with 38.2 BLEU@4 and 126.2 CIDEr, demonstrating exceptional performance without pre-training or ancillary modules. This indicates strong competitiveness with AR models, revealing the previously untapped potential of diffusion models in image-to-text generation.

Video paragraph captioning (VPC) involves generating detailed narratives for long videos, utilizing supportive modalities such as speech and event boundaries. However, the existing models are constrained by the assumption of constant availability of a single auxiliary modality, which is impractical given the diversity and unpredictable nature of real-world scenarios. To this end, we propose a Missing-Resistant framework MR-VPC that effectively harnesses all available auxiliary inputs and maintains resilience even in the absence of certain modalities. Under this framework, we propose the Multimodal VPC (MVPC) architecture integrating video, speech, and event boundary inputs in a unified manner to process various auxiliary inputs. Moreover, to fortify the model against incomplete data, we introduce DropAM, a data augmentation strategy that randomly omits auxiliary inputs, paired with DistillAM, a regularization target that distills knowledge from teacher models trained on modality-complete data, enabling efficient learning in modality-deficient environments. Through exhaustive experimentation on YouCook2 and ActivityNet Captions, MR-VPC has proven to deliver superior performance on modality-complete and modality-missing test data. This work highlights the significance of developing resilient VPC models and paves the way for more adaptive, robust multimodal video understanding., Comment: Code available at https://github.com/lancopku/MR-VPC

Financial prediction from Monetary Policy Conference (MPC) calls is a new yet challenging task, which targets at predicting the price movement and volatility for specific financial assets by analyzing multimodal information including text, video, and audio. Although the existing work has achieved great success using cross-modal transformer blocks, it overlooks the potential external financial knowledge, the varying contributions of different modalities to financial prediction, as well as the innate relations among different financial assets. To tackle these limitations, we propose a novel Modal-Adaptive kNowledge-enhAnced Graph-basEd financial pRediction scheme, named MANAGER. Specifically, MANAGER resorts to FinDKG to obtain the external related knowledge for the input text. Meanwhile, MANAGER adopts BEiT-3 and Hidden-unit BERT (HuBERT) to extract the video and audio features, respectively. Thereafter, MANAGER introduces a novel knowledge-enhanced cross-modal graph that fully characterizes the semantic relations among text, external knowledge, video and audio, to adaptively utilize the information in different modalities, with ChatGLM2 as the backbone. Extensive experiments on a publicly available dataset Monopoly verify the superiority of our model over cutting-edge methods., Comment: Accepted by LREC Coling 2024 -FinNLP (oral)

Recently, there is a surge in interest surrounding video large language models (Video LLMs). However, existing benchmarks fail to provide a comprehensive feedback on the temporal perception ability of Video LLMs. On the one hand, most of them are unable to distinguish between different temporal aspects (e.g., speed, direction) and thus cannot reflect the nuanced performance on these specific aspects. On the other hand, they are limited in the diversity of task formats (e.g., only multi-choice QA), which hinders the understanding of how temporal perception performance may vary across different types of tasks. Motivated by these two problems, we propose the \textbf{TempCompass} benchmark, which introduces a diversity of temporal aspects and task formats. To collect high-quality test data, we devise two novel strategies: (1) In video collection, we construct conflicting videos that share the same static content but differ in a specific temporal aspect, which prevents Video LLMs from leveraging single-frame bias or language priors. (2) To collect the task instructions, we propose a paradigm where humans first annotate meta-information for a video and then an LLM generates the instruction. We also design an LLM-based approach to automatically and accurately evaluate the responses from Video LLMs. Based on TempCompass, we comprehensively evaluate 8 state-of-the-art (SOTA) Video LLMs and 3 Image LLMs, and reveal the discerning fact that these models exhibit notably poor temporal perception ability. Our data will be available at https://github.com/llyx97/TempCompass.

Driven by the rapid development of Large Language Models (LLMs), LLM-based agents have been developed to handle various real-world applications, including finance, healthcare, and shopping, etc. It is crucial to ensure the reliability and security of LLM-based agents during applications. However, the safety issues of LLM-based agents are currently under-explored. In this work, we take the first step to investigate one of the typical safety threats, backdoor attack, to LLM-based agents. We first formulate a general framework of agent backdoor attacks, then we present a thorough analysis of different forms of agent backdoor attacks. Specifically, compared with traditional backdoor attacks on LLMs that are only able to manipulate the user inputs and model outputs, agent backdoor attacks exhibit more diverse and covert forms: (1) From the perspective of the final attacking outcomes, the agent backdoor attacker can not only choose to manipulate the final output distribution, but also introduce the malicious behavior in an intermediate reasoning step only, while keeping the final output correct. (2) Furthermore, the former category can be divided into two subcategories based on trigger locations, in which the backdoor trigger can either be hidden in the user query or appear in an intermediate observation returned by the external environment. We implement the above variations of agent backdoor attacks on two typical agent tasks including web shopping and tool utilization. Extensive experiments show that LLM-based agents suffer severely from backdoor attacks and such backdoor vulnerability cannot be easily mitigated by current textual backdoor defense algorithms. This indicates an urgent need for further research on the development of targeted defenses against backdoor attacks on LLM-based agents. Warning: This paper may contain biased content., Comment: Accepted at NeurIPS 2024, camera ready version. Code and data are available at https://github.com/lancopku/agent-backdoor-attacks

This work proposes TimeChat, a time-sensitive multimodal large language model specifically designed for long video understanding. Our model incorporates two key architectural contributions: (1) a timestamp-aware frame encoder that binds visual content with the timestamp of each frame, and (2) a sliding video Q-Former that produces a video token sequence of varying lengths to accommodate videos of various durations. Additionally, we construct an instruction-tuning dataset, encompassing 6 tasks and a total of 125K instances, to further enhance TimeChat's instruction-following performance. Experiment results across various video understanding tasks, such as dense captioning, temporal grounding, and highlight detection, demonstrate TimeChat's strong zero-shot temporal localization and reasoning capabilities. For example, it achieves +9.2 F1 score and +2.8 CIDEr on YouCook2, +5.8 HIT@1 on QVHighlights, and +27.5 R@1 (IoU=0.5) on Charades-STA, compared to state-of-the-art video large language models, holding the potential to serve as a versatile video assistant for long-form video comprehension tasks and satisfy realistic user requirements., Comment: CVPR 2024 camera-ready version, code is available at https://github.com/RenShuhuai-Andy/TimeChat