Search

Multipartite quantum scenarios are a significant and challenging resource in quantum information science. Tensors provide a powerful framework for representing multipartite quantum systems. In this work, we introduce the role of tensor-generated matrices that can broadly be defined as the relationships between an $m$-th order $n$-dimensional tensor and an $n$-dimensional square matrix. Through these established connections, we demonstrate that the classification of the tensor-generated matrix as an $H$-matrix implies the original tensor is also an $H$-tensor. We also explore various similar properties exhibited by both the original tensor and the tensor-generated matrix, including weak irreducibility, weakly chained diagonal dominance, and (strong) symmetry. These findings provide a method to transform intricate tensor problems into matrices in specific contexts, which is especially pertinent due to the NP-hard complexity of the majority of tensor problems. Subsequently, we explore the application of tensor-generated matrices in analyzing the classicality of spin states. Leveraging the tensor representation, we introduce classicality criteria for (strongly) symmetric spin-$j$ states, which potentially provide fresh perspectives on the study of multipartite quantum resources. Finally, we extend classical matrix eigenvalue inclusion sets to higher-order tensor $H$-eigenvalues, a task that is typically challenging for higher-order tensors. Consequently, we propose representative tensor $H$-eigenvalue inclusion sets, such as modified Brauer's Ovals of Cassini sets, Ostrowski sets, and $S$-type inclusion sets.

We introduce the Principled Reasoning and Acting (PRAct) framework, a novel method for learning and enforcing action principles from trajectory data. Central to our approach is the use of text gradients from a reflection and optimization engine to derive these action principles. To adapt action principles to specific task requirements, we propose a new optimization framework, Reflective Principle Optimization (RPO). After execution, RPO employs a reflector to critique current action principles and an optimizer to update them accordingly. We develop the RPO framework under two scenarios: Reward-RPO, which uses environmental rewards for reflection, and Self-RPO, which conducts self-reflection without external rewards. Additionally, two RPO methods, RPO-Traj and RPO-Batch, is introduced to adapt to different settings. Experimental results across four environments demonstrate that the PRAct agent, leveraging the RPO framework, effectively learns and applies action principles to enhance performance., Comment: Accepted to SIG CoNLL 2024

TIC 157365951 has been classified as a $\delta$ Scuti type by the International Variable Star Index (VSX). Through the spectra from Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) and its light curve, we further discovered that it is a binary system. This binary system comprises a red dwarf star and a compact star. Through the spectral energy distribution (SED) fitting, we determined the mass of the red dwarf star as $M_1 = 0.31 \pm 0.01 M_{\odot}$ and its radius as $R_1 = 0.414 \pm 0.004 R_{\odot}$. By fitting the double-peaked H${\rm \alpha}$ emission, we derived the mass ratio of $q = 1.76 \pm 0.04 $, indicating a compact star mass of $M_2 = 0.54 \pm 0.01 M_{\odot}$. Using Phoebe to model the light curve and radial velocity curve for the detached binary system, we obtained a red dwarf star mass of $M_1 = 0.29 \pm 0.02 M_{\odot}$, a radius of $R_1 = 0.39 \pm 0.04 R_{\odot}$, and a Roche-lobe filling factor of $f = 0.995\pm0.129$, which is close to the $f=1$ expected for a semi-detached system. The Phoebe model gives a compact star mass $M_2 = 0.53 \pm 0.05 M_{\odot}$. Constraining the system to be semidetached gives $M_1 = 0.34 \pm 0.02 M_{\odot}$, $R_1 = 0.41 \pm 0.01 R_{\odot}$, and $M_2 = 0.62 \pm 0.03 M_{\odot}$. The consistency of the models is encouraging. The value of the Roche-lobe filling factor suggests that there might be ongoing mass transfer. The compact star mass is as massive as a typical white dwarf.

Recent advancements in transition metal dichalcogenides (TMDs) have unveiled exceptional optical and electronic characteristics, opened up new opportunities, and provided a unique platform for exploring light-matter interactions under the strong coupling regime. The exploitation of exciton-polaritons, with their peculiar hybrid light-matter properties, for the development of spintronic customizable devices that enhance both the information capacity and functionality at ambient temperatures is often suggested as a promising route. However, although TMD polaritons have shown promising potential, the microscopic mechanisms leading to nonlinearities in TMD polaritons are complex and their spin-anisotropy, a crucial requirement for many proposed polaritonic devices, has been missing. Here, we demonstrate the absence of spin-anisotropic interaction in a monolayer WS2 microcavity (at room temperature) and show how spin-dependent interactions can be controlled and spin anisotropy recovered by engineering double WS2 layer structures with varied interlayer spacing. We attribute this phenomenon to a distinctive feature in exciton-polariton physics: layer-dependent polariton-phonon coupling. We use theoretical calculations of the phonon electrostatic potentials finding a drastically different coupling strength for single and double monolayer samples and discuss qualitatively how this explains the observed spin-anisotropic response. This is further consistent with experiments on multi WS2 layer samples and the identification of a critical separation distance, above which an effective single monolayer spin-anisotropic response is recovered, both in experiment and theory. Our work lays the groundwork for the development of spin-optronic polaritonic devices at room temperature.

Key-Value (KV) caching is a common technique to enhance the computational efficiency of Large Language Models (LLMs), but its memory overhead grows rapidly with input length. Prior work has shown that not all tokens are equally important for text generation, proposing layer-level KV cache compression to selectively retain key information. Recognizing the distinct roles of attention heads in generation, we propose HeadKV, a head-level KV cache compression method, and HeadKV-R2, which leverages a novel contextual reasoning ability estimation for compression. Our approach operates at the level of individual heads, estimating their importance for contextual QA tasks that require both retrieval and reasoning capabilities. Extensive experiments across diverse benchmarks (LongBench, LooGLE), model architectures (e.g., Llama-3-8B-Instruct, Mistral-7B-Instruct), and long-context abilities tests demonstrate that our head-level KV cache compression significantly outperforms strong baselines, particularly in low-resource settings (KV size = 64 & 128). Notably, our method retains just 1.5% of the KV cache while achieving 97% of the performance of the full KV cache on the contextual question answering benchmark.Codes are available at https://github.com/FYYFU/HeadKV, Comment: 18pages

Bound states in the continuum (BICs) are quantum states that remain localized despite existing within a continuum of extended, delocalized states. They defy conventional wave theories and could be instrumental for quantum technologies that rely on the precise control of quantum states. While optical BICs have been realized in photonic systems, achieving electronic bound states in a metallic background remains an ongoing challenge. Here, we observe two defect states that remain localized within the metallic continuum of Pd5AlI2, a two-dimensional van der Waals metal. The emergence of these states is a manifestation of the hopping interference in the Pd5AlI2 lattice. This interference results in a (quasi) flat band and spatially localized eigenstates that are orthogonal to the metallic continuum thus avoiding hybridization with extended states.

Large language models (LLMs) represent a transformative class of AI tools capable of revolutionizing various aspects of healthcare by generating human-like responses across diverse contexts and adapting to novel tasks following human instructions. Their potential application spans a broad range of medical tasks, such as clinical documentation, matching patients to clinical trials, and answering medical questions. In this primer paper, we propose an actionable guideline to help healthcare professionals more efficiently utilize LLMs in their work, along with a set of best practices. This approach consists of several main phases, including formulating the task, choosing LLMs, prompt engineering, fine-tuning, and deployment. We start with the discussion of critical considerations in identifying healthcare tasks that align with the core capabilities of LLMs and selecting models based on the selected task and data, performance requirements, and model interface. We then review the strategies, such as prompt engineering and fine-tuning, to adapt standard LLMs to specialized medical tasks. Deployment considerations, including regulatory compliance, ethical guidelines, and continuous monitoring for fairness and bias, are also discussed. By providing a structured step-by-step methodology, this tutorial aims to equip healthcare professionals with the tools necessary to effectively integrate LLMs into clinical practice, ensuring that these powerful technologies are applied in a safe, reliable, and impactful manner., Comment: Under review

The present article is an empirical study that investigates the learning situation of linear algebra. Research was performed among 60 science and engineering students from different universities in Zhejiang, Jiangsu, Hubei, and Shandong who were taught by three mathematics teachers in the course of a 3-day short-term training consisting of 24 classes; following this, the effectiveness of short-term training for mathematics curriculum learning in higher education was evaluated based on the IRT measurement model. The results indicate that regardless of the difficulty of the test, short-term training significantly enhances students' mastery of linear algebra knowledge points, and students can accurately perceive their mathematical achievements; the effect observed is not consistent with that of private tutoring (PT) in primary and secondary schools., Comment: 39 pages. comments are welcome

Reinforcement Learning from Human Feedback (RLHF) has proven highly effective in aligning Large Language Models (LLMs) with human preferences. However, the original RLHF typically optimizes under an overall reward, which can lead to a suboptimal learning process. This limitation stems from RLHF's lack of awareness regarding which specific tokens should be reinforced or suppressed. Moreover, conflicts in supervision can arise, for instance, when a chosen response includes erroneous tokens, while a rejected response contains accurate elements. To rectify these shortcomings, increasing dense reward methods, such as step-wise and token-wise RLHF, have been proposed. However, these existing methods are limited to specific tasks (like mathematics). In this paper, we propose the ``Adaptive Message-wise RLHF'' method, which robustly applies to various tasks. By defining pivot tokens as key indicators, our approach adaptively identifies essential information and converts sample-level supervision into fine-grained, subsequence-level supervision. This aligns the density of rewards and action spaces more closely with the information density of the input. Experiments demonstrate that our method can be integrated into various training methods, significantly mitigating hallucinations and catastrophic forgetting problems while outperforming other methods on multiple evaluation metrics. Our method improves the success rate on adversarial samples by 10\% compared to the sample-wise approach and achieves a 1.3\% improvement on evaluation benchmarks such as MMLU, GSM8K, and HumanEval et al.

LEIA (Lobster Eye Imager for Astronomy) detected a new X-ray transient on November 7, 2022, identified as a superflare event occurring on a nearby RS CVn-type binary HD 251108. The flux increase was also detected in follow-up observations at X-ray, UV and optical wavelengths. The flare lasted for about 40 days in soft X-ray observations, reaching a peak luminosity of ~1.1 * 10^34 erg/s in 0.5-4.0 keV, which is roughly 60 times the quiescent luminosity. Optical brightening was observed for only one night. The X-ray light curve is well described by a double "FRED" (fast rise and exponential decay) model, attributed to the cooling process of a loop arcade structure formed subsequent to the initial large loop with a half-length of ~1.9 times the radius of the host star. Time-resolved X-ray spectra were fitted with a two-temperature apec model, showing significant evolution of plasma temperature, emission measure, and metal abundance over time. The estimated energy released in the LEIA band is ~3 * 10^39 erg, suggesting this is likely the most energetic X-ray stellar flare with the longest duration detected to date., Comment: submitted to ApJL, 22 pages, 9 figures, 7 tables

Index tuning is crucial for optimizing database performance by selecting optimal indexes based on workload. The key to this process lies in an accurate and efficient benefit estimator. Traditional methods relying on what-if tools often suffer from inefficiency and inaccuracy. In contrast, learning-based models provide a promising alternative but face challenges such as instability, lack of interpretability, and complex management. To overcome these limitations, we adopt a novel approach: quantifying the uncertainty in learning-based models' results, thereby combining the strengths of both traditional and learning-based methods for reliable index tuning. We propose Beauty, the first uncertainty-aware framework that enhances learning-based models with uncertainty quantification and uses what-if tools as a complementary mechanism to improve reliability and reduce management complexity. Specifically, we introduce a novel method that combines AutoEncoder and Monte Carlo Dropout to jointly quantify uncertainty, tailored to the characteristics of benefit estimation tasks. In experiments involving sixteen models, our approach outperformed existing uncertainty quantification methods in the majority of cases. We also conducted index tuning tests on six datasets. By applying the Beauty framework, we eliminated worst-case scenarios and more than tripled the occurrence of best-case scenarios., Comment: 14 pages, 11 figures

Altermagnetism exhibits advantages over both ferromagnetic and antiferromagnetic counterparts by enabling spin splitting within antiferromagnetic materials. Currently, it is established that valley polarization in altermagnetism remains largely insensitive to spin-orbit coupling and spin. Here, using Cr$_2$S$_2$ as a case study, we investigate the mechanism through which an external field modulates valley polarization in altermagnetism. This effect arises from the external field's disruption of diagonal mirror symmetry $M_{xy}$, consequently inducing valley polarization within the material. Strain not only induces valley polarization but also generates an almost uniform magnetic field, which can reach as high as 118.39 T under 5% uniaxial strain. In addition, this symmetry breaking in Cr$_2$S$_2$ monolayers results in significant piezomagnetic properties, merging piezomagnetic and altermagnetic characteristics in two-dimensional materials.

Since the discovery of superconductivity in the Fe(Te,Se) system, it has been a general consensus that the end member of FeTe is not superconducting. Nonetheless, in recent years, there have been reports of superconducting FeTe films, but the origin of their superconductivity remains mysterious. Here, we provide the first comprehensive review of all the reported FeTe films regarding the relationship between their superconductivity and neighboring layers. Based on this review, we show that telluride neighboring layers are the key to superconducting FeTe films. Then, with additional new studies, we show that stoichiometric Te content, which can be readily achieved in FeTe films with the assistance of neighboring telluride layers, might be crucial to stabilizing the superconductivity in this system. This work provides insights into the underlying mechanism behind superconductivity in FeTe films and sheds light on the critical role of neighboring layers and stoichiometry control toward manipulating topological superconductivity in FeTe heterostructures., Comment: 16 pages, 4 figures

Visual preference alignment involves training Large Vision-Language Models (LVLMs) to predict human preferences between visual inputs. This is typically achieved by using labeled datasets of chosen/rejected pairs and employing optimization algorithms like direct preference optimization (DPO). Existing visual alignment methods, primarily designed for single-image scenarios, struggle to effectively handle the complexity of multi-image tasks due to the scarcity of diverse training data and the high cost of annotating chosen/rejected pairs. We present Multi-Image Augmented Direct Preference Optimization (MIA-DPO), a visual preference alignment approach that effectively handles multi-image inputs. MIA-DPO mitigates the scarcity of diverse multi-image training data by extending single-image data with unrelated images arranged in grid collages or pic-in-pic formats, significantly reducing the costs associated with multi-image data annotations. Our observation reveals that attention values of LVLMs vary considerably across different images. We use attention values to identify and filter out rejected responses the model may have mistakenly focused on. Our attention-aware selection for constructing the chosen/rejected pairs without relying on (i) human annotation, (ii) extra data, and (iii) external models or APIs. MIA-DPO is compatible with various architectures and outperforms existing methods on five multi-image benchmarks, achieving an average performance boost of 3.0% on LLaVA-v1.5 and 4.3% on the recent InternLM-XC2.5. Moreover, MIA-DPO has a minimal effect on the model's ability to understand single images., Comment: Project URL: https://github.com/Liuziyu77/MIA-DPO

Superconductors, an essential class of functional materials, hold a vital position in both fundamental science and practical applications. However, most superconductors, including MgB$_2$, Bi$_2$Sr$_2$CaCu$_2$O$_{8+\delta}$, and FeSe, are highly sensitive to environmental attacks (such as water and moist air), hindering their wide applications. More importantly, the surface physical and chemical processes of most superconductors in various environments remain poorly understood. Here, we comprehensively investigate the high resistance of superconducting titanium nitride (TiN) epitaxial films against acid and alkali attacks. Unexpectedly, despite immersion in acid and alkaline solutions for over 7 days, the crystal structure and superconducting properties of TiN films remain stable, as demonstrated by high-resolution X-ray diffraction, electrical transport, atomic force microscopy, and scanning electron microscope. Furthermore, combining scanning transmission electron microscopy analysis with density functional theory calculations revealed the corrosion mechanisms: acid corrosions lead to the creation of numerous defects due to the substitution of Cl ions for N anions, whereas alkaline environments significantly reduce the film thickness through the stabilization of OH$^\ast$ adsorbates. Our results uncover the unexpected stability and durability of superconducting materials against environmental attacks, highlighting their potential for enhanced reliability and longevity in diverse applications., Comment: 10 pages, 8 figures

Causal reasoning (CR) is a crucial aspect of intelligence, essential for problem-solving, decision-making, and understanding the world. While large language models (LLMs) can generate rationales for their outputs, their ability to reliably perform causal reasoning remains uncertain, often falling short in tasks requiring a deep understanding of causality. In this survey, we provide a comprehensive review of research aimed at enhancing LLMs for causal reasoning. We categorize existing methods based on the role of LLMs: either as reasoning engines or as helpers providing knowledge or data to traditional CR methods, followed by a detailed discussion of the methodologies in each category. We then evaluate the performance of LLMs on various causal reasoning tasks, providing key findings and in-depth analysis. Finally, we provide insights from current studies and highlight promising directions for future research. We aim for this work to serve as a comprehensive resource, fostering further advancements in causal reasoning with LLMs. Resources are available at https://github.com/chendl02/Awesome-LLM-causal-reasoning.

Multimodal Large Language Models (MLLMs) have shown promising progress in understanding and analyzing video content. However, processing long videos remains a significant challenge constrained by LLM's context size. To address this limitation, we propose LongVU, a spatiotemporal adaptive compression mechanism thats reduces the number of video tokens while preserving visual details of long videos. Our idea is based on leveraging cross-modal query and inter-frame dependencies to adaptively reduce temporal and spatial redundancy in videos. Specifically, we leverage DINOv2 features to remove redundant frames that exhibit high similarity. Then we utilize text-guided cross-modal query for selective frame feature reduction. Further, we perform spatial token reduction across frames based on their temporal dependencies. Our adaptive compression strategy effectively processes a large number of frames with little visual information loss within given context length. Our LongVU consistently surpass existing methods across a variety of video understanding benchmarks, especially on hour-long video understanding tasks such as VideoMME and MLVU. Given a light-weight LLM, our LongVU also scales effectively into a smaller size with state-of-the-art video understanding performance., Comment: Project page: https://vision-cair.github.io/LongVU

Zero-Shot Composed Image Retrieval (ZS-CIR) supports diverse tasks with a broad range of visual content manipulation intentions that can be related to domain, scene, object, and attribute. A key challenge for ZS-CIR is to accurately map image representation to a pseudo-word token that captures the manipulation intention relevant image information for generalized CIR. However, existing methods between the retrieval and pre-training stages lead to significant redundancy in the pseudo-word tokens. In this paper, we propose a novel denoising image-to-word mapping approach, named Denoise-I2W, for mapping images into denoising pseudo-word tokens that, without intention-irrelevant visual information, enhance accurate ZS-CIR. Specifically, a pseudo triplet construction module first automatically constructs pseudo triples (\textit{i.e.,} a pseudo-reference image, a pseudo-manipulation text, and a target image) for pre-training the denoising mapping network. Then, a pseudo-composed mapping module maps the pseudo-reference image to a pseudo-word token and combines it with the pseudo-manipulation text with manipulation intention. This combination aligns with the target image, facilitating denoising intention-irrelevant visual information for mapping. Our proposed Denoise-I2W is a model-agnostic and annotation-free approach. It demonstrates strong generalization capabilities across three state-of-the-art ZS-CIR models on four benchmark datasets. By integrating Denoise-I2W with existing best models, we obtain consistent and significant performance boosts ranging from 1.45\% to 4.17\% over the best methods without increasing inference costs. and achieve new state-of-the-art results on ZS-CIR. Our code is available at \url{https://github.com/Pter61/denoise-i2w-tmm}., Comment: This work was submitted to IJCAI 2024, with a score of weak accept and borderline accept

In this study, we present the pulse profile of the unique and the second brightest gamma-ray burst GRB 230307A, and analyze its temporal behavior using a joint GECAM--Fermi/GBM time-resolved spectral analysis. The utilization of GECAM data is advantageous as it successfully captured significant data during the pile-up period of the Fermi/GBM. We investigate the evolution of its flux, photon fluence, photon flux, peak energy, and the corresponding hardness-intensity and hardness-flux correlations. The findings within the first 27 seconds exhibit consistent patterns reported previously, providing valuable insights for comparing observations with predictions from the synchrotron radiation model invoking an expanding shell. Beyond the initial 27 seconds, we observe a notable transition in the emitted radiation, attributed to high latitude emission (HLE), influenced by the geometric properties of the shells and the relativistic Doppler effects. By modeling the data within the framework of the large-radius internal shock model, we discuss the required parameters as well as the limitations of the model. We conclude that a more complicated synchrotron emission model is needed to fully describe the observational data of GRB 230307A., Comment: Accepted for publication in The Astrophysical Journal (ApJ)

Short-form video content has become increasingly popular and influential in recent years. Its concise yet engaging format aligns well with todays' fast-paced and on-the-go lifestyles, making it a dominating trend in the digital world. As one of the front runners in the short video platform space, ByteDance has been highly successful in delivering a one-of-a-kind short video experience and attracting billions of users worldwide. One key contributing factor is its advanced end-to-end personalized short video playback technology, where we pioneered and developed the new technical field over the past five years to optimize user experience. This paper introduces the major concepts and methodologies of this personalized video playback technology that distinguish it from traditional multimedia technologies. More details, including goal setting, iterative process, modeling, experimental methods and required supporting systems, are also provided to encourage deeper research in this area.

Knowledge graphs (KGs) generated by large language models (LLMs) are becoming increasingly valuable for Retrieval-Augmented Generation (RAG) applications that require knowledge-intensive reasoning. However, existing KG extraction methods predominantly rely on prompt-based approaches, which are inefficient for processing large-scale corpora. These approaches often suffer from information loss, particularly with long documents, due to the lack of specialized design for KG construction. Additionally, there is a gap in evaluation datasets and methodologies for ontology-free KG construction. To overcome these limitations, we propose SynthKG, a multi-step, document-level ontology-free KG synthesis workflow based on LLMs. By fine-tuning a smaller LLM on the synthesized document-KG pairs, we streamline the multi-step process into a single-step KG generation approach called Distill-SynthKG, substantially reducing the number of LLM inference calls. Furthermore, we re-purpose existing question-answering datasets to establish KG evaluation datasets and introduce new evaluation metrics. Using KGs produced by Distill-SynthKG, we also design a novel graph-based retrieval framework for RAG. Experimental results demonstrate that Distill-SynthKG not only surpasses all baseline models in KG quality -- including models up to eight times larger -- but also consistently excels in retrieval and question-answering tasks. Our proposed graph retrieval framework also outperforms all KG-retrieval methods across multiple benchmark datasets. We release the SynthKG dataset and Distill-SynthKG model publicly to support further research and development.

Retrieval module can be plugged into many downstream NLP tasks to improve their performance, such as open-domain question answering and retrieval-augmented generation. The key to a retrieval system is to calculate relevance scores to query and passage pairs. However, the definition of relevance is often ambiguous. We observed that a major class of relevance aligns with the concept of entailment in NLI tasks. Based on this observation, we designed a method called entailment tuning to improve the embedding of dense retrievers. Specifically, we unify the form of retrieval data and NLI data using existence claim as a bridge. Then, we train retrievers to predict the claims entailed in a passage with a variant task of masked prediction. Our method can be efficiently plugged into current dense retrieval methods, and experiments show the effectiveness of our method., Comment: EMNLP 2024 Main

Solving constrained nonlinear programs (NLPs) is of great importance in various domains such as power systems, robotics, and wireless communication networks. One widely used approach for addressing NLPs is the interior point method (IPM). The most computationally expensive procedure in IPMs is to solve systems of linear equations via matrix factorization. Recently, machine learning techniques have been adopted to expedite classic optimization algorithms. In this work, we propose using Long Short-Term Memory (LSTM) neural networks to approximate the solution of linear systems and integrate this approximating step into an IPM. The resulting approximate NLP solution is then utilized to warm-start an interior point solver. Experiments on various types of NLPs, including Quadratic Programs and Quadratically Constrained Quadratic Programs, show that our approach can significantly accelerate NLP solving, reducing iterations by up to 60% and solution time by up to 70% compared to the default solver., Comment: Accepted by NeurIPS 2024

We present xGen-MM-Vid (BLIP-3-Video): a multimodal language model for videos, particularly designed to efficiently capture temporal information over multiple frames. BLIP-3-Video takes advantage of the 'temporal encoder' in addition to the conventional visual tokenizer, which maps a sequence of tokens over multiple frames into a compact set of visual tokens. This enables BLIP3-Video to use much fewer visual tokens than its competing models (e.g., 32 vs. 4608 tokens). We explore different types of temporal encoders, including learnable spatio-temporal pooling as well as sequential models like Token Turing Machines. We experimentally confirm that BLIP-3-Video obtains video question-answering accuracies comparable to much larger state-of-the-art models (e.g., 34B), while being much smaller (i.e., 4B) and more efficient by using fewer visual tokens. The project website is at https://www.salesforceairesearch.com/opensource/xGen-MM-Vid/index.html

Large language models (LLM) have been attracting much attention from the community recently, due to their remarkable performance in all kinds of downstream tasks. According to the well-known scaling law, scaling up a dense LLM enhances its capabilities, but also significantly increases the computational complexity. Mixture-of-Experts (MoE) models address that by allowing the model size to grow without substantially raising training or inference costs. Yet MoE models face challenges regarding knowledge sharing among experts, making their performance somehow sensitive to routing accuracy. To tackle that, previous works introduced shared experts and combined their outputs with those of the top $K$ routed experts in an ``addition'' manner. In this paper, inspired by collective matrix factorization to learn shared knowledge among data, we propose CartesianMoE, which implements more effective knowledge sharing among experts in more like a ``multiplication'' manner. Extensive experimental results indicate that CartesianMoE outperforms previous MoE models for building LLMs, in terms of both perplexity and downstream task performance. And we also find that CartesianMoE achieves better expert routing robustness.

Despite the remarkable success of deep neural networks (DNNs) in computer vision, they fail to remain high-performing when facing distribution shifts between training and testing data. In this paper, we propose Knowledge-Guided Visual representation learning (KGV), a distribution-based learning approach leveraging multi-modal prior knowledge, to improve generalization under distribution shift. We use prior knowledge from two distinct modalities: 1) a knowledge graph (KG) with hierarchical and association relationships; and 2) generated synthetic images of visual elements semantically represented in the KG. The respective embeddings are generated from the given modalities in a common latent space, i.e., visual embeddings from original and synthetic images as well as knowledge graph embeddings (KGEs). These embeddings are aligned via a novel variant of translation-based KGE methods, where the node and relation embeddings of the KG are modeled as Gaussian distributions and translations respectively. We claim that incorporating multi-model prior knowledge enables more regularized learning of image representations. Thus, the models are able to better generalize across different data distributions. We evaluate KGV on different image classification tasks with major or minor distribution shifts, namely road sign classification across datasets from Germany, China, and Russia, image classification with the mini-ImageNet dataset and its variants, as well as the DVM-CAR dataset. The results demonstrate that KGV consistently exhibits higher accuracy and data efficiency than the baselines across all experiments.

Recent large visual-language action models pretrained on diverse robot datasets have demonstrated the potential for generalizing to new environments with a few in-domain data. However, those approaches usually predict discretized or continuous actions by a small action head, which limits the ability in handling diverse action spaces. In contrast, we model the continuous action with a large multi-modal diffusion transformer, dubbed as Diffusion Transformer Policy, in which we directly denoise action chunks by a large transformer model rather than a small action head. By leveraging the scaling capability of transformers, the proposed approach can effectively model continuous end-effector actions across large diverse robot datasets, and achieve better generalization performance. Extensive experiments demonstrate Diffusion Transformer Policy pretrained on diverse robot data can generalize to different embodiments, including simulation environments like Maniskill2 and Calvin, as well as the real-world Franka arm. Specifically, without bells and whistles, the proposed approach achieves state-of-the-art performance with only a single third-view camera stream in the Calvin novel task setting (ABC->D), improving the average number of tasks completed in a row of 5 to 3.6, and the pretraining stage significantly facilitates the success sequence length on the Calvin by over 1.2. The code will be publicly available., Comment: Preprint

Recovering a spectrum of diverse policies from a set of expert trajectories is an important research topic in imitation learning. After determining a latent style for a trajectory, previous diverse policies recovering methods usually employ a vanilla behavioral cloning learning objective conditioned on the latent style, treating each state-action pair in the trajectory with equal importance. Based on an observation that in many scenarios, behavioral styles are often highly relevant with only a subset of state-action pairs, this paper presents a new principled method in diverse polices recovery. In particular, after inferring or assigning a latent style for a trajectory, we enhance the vanilla behavioral cloning by incorporating a weighting mechanism based on pointwise mutual information. This additional weighting reflects the significance of each state-action pair's contribution to learning the style, thus allowing our method to focus on state-action pairs most representative of that style. We provide theoretical justifications for our new objective, and extensive empirical evaluations confirm the effectiveness of our method in recovering diverse policies from expert data., Comment: 18 pages, 6 figures

Evaluating retrieval-augmented generation (RAG) systems remains challenging, particularly for open-ended questions that lack definitive answers and require coverage of multiple sub-topics. In this paper, we introduce a novel evaluation framework based on sub-question coverage, which measures how well a RAG system addresses different facets of a question. We propose decomposing questions into sub-questions and classifying them into three types -- core, background, and follow-up -- to reflect their roles and importance. Using this categorization, we introduce a fine-grained evaluation protocol that provides insights into the retrieval and generation characteristics of RAG systems, including three commercial generative answer engines: You.com, Perplexity AI, and Bing Chat. Interestingly, we find that while all answer engines cover core sub-questions more often than background or follow-up ones, they still miss around 50% of core sub-questions, revealing clear opportunities for improvement. Further, sub-question coverage metrics prove effective for ranking responses, achieving 82% accuracy compared to human preference annotations. Lastly, we also demonstrate that leveraging core sub-questions enhances both retrieval and answer generation in a RAG system, resulting in a 74% win rate over the baseline that lacks sub-questions.

The rising need for explainable deep neural network architectures has utilized semantic concepts as explainable units. Several approaches utilizing disentangled representation learning estimate the generative factors and utilize them as concepts for explaining DNNs. However, even though the generative factors for a dataset remain fixed, concepts are not fixed entities and vary based on downstream tasks. In this paper, we propose a disentanglement mechanism utilizing a variational autoencoder (VAE) for learning mutually independent generative factors for a given dataset and subsequently learning task-specific concepts using a structural causal model (SCM). Our method assumes generative factors and concepts to form a bipartite graph, with directed causal edges from generative factors to concepts. Experiments are conducted on datasets with known generative factors: D-sprites and Shapes3D. On specific downstream tasks, our proposed method successfully learns task-specific concepts which are explained well by the causal edges from the generative factors. Lastly, separate from current causal concept discovery methods, our methodology is generalizable to an arbitrary number of concepts and flexible to any downstream tasks.

Previous studies have highlighted significant advancements in multimodal fusion. Nevertheless, such methods often encounter challenges regarding the efficacy of feature extraction, data integrity, consistency of feature dimensions, and adaptability across various downstream tasks. This paper proposes a generalized multimodal fusion method (GMF) via the Poisson-Nernst-Planck (PNP) equation, which adeptly addresses the aforementioned issues. Theoretically, the optimization objective for traditional multimodal tasks is formulated and redefined by integrating information entropy and the flow of gradient backward step. Leveraging these theoretical insights, the PNP equation is applied to feature fusion, rethinking multimodal features through the framework of charged particles in physics and controlling their movement through dissociation, concentration, and reconstruction. Building on these theoretical foundations, GMF disassociated features which extracted by the unimodal feature extractor into modality-specific and modality-invariant subspaces, thereby reducing mutual information and subsequently lowering the entropy of downstream tasks. The identifiability of the feature's origin enables our approach to function independently as a frontend, seamlessly integrated with a simple concatenation backend, or serve as a prerequisite for other modules. Experimental results on multiple downstream tasks show that the proposed GMF achieves performance close to the state-of-the-art (SOTA) accuracy while utilizing fewer parameters and computational resources. Furthermore, by integrating GMF with advanced fusion methods, we surpass the SOTA results., Comment: NeurIPS 2024 Rejected paper, 28 pages

The Directed Feedback Vertex Set problem (DFVS) asks whether it is possible to remove at most $k$ vertices from a directed graph to make it acyclic. Whether DFVS is fixed-parameter tractable was a long-standing open problem in parameterized complexity until it was solved by Chen et al. in 2008 (STOC 2008). Now the running-time bound of this problem is improved to $\mathcal O(k!4^kk^5(n+m))$ (Lokshtanov et al, SODA 2018), where $n$ and $m$ are the numbers of vertices and arcs in the graph. In this paper, we simplify one crucial step in all previous parameterized algorithms for DFVS, which is to solve the compression version of the problem, and refine the running-time bound for DFVS to $\mathcal O(k!2^{o(k)}(n+m))$.

Representation learning of Text-Attributed Graphs (TAGs) has garnered significant attention due to its applications in various domains, including recommendation systems and social networks. Despite advancements in TAG learning methodologies, challenges remain in explainability due to the black-box nature of existing TAG representation learning models. This paper presents TAGExplainer, the first method designed to generate natural language explanations for TAG learning. TAGExplainer employs a generative language model that maps input-output pairs to explanations reflecting the model's decision-making process. To address the lack of annotated ground truth explanations in real-world scenarios, we propose first generating pseudo-labels that capture the model's decisions from saliency-based explanations, then the pseudo-label generator is iteratively trained based on three training objectives focusing on faithfulness and brevity via Expert Iteration, to improve the quality of generated pseudo-labels. The high-quality pseudo-labels are finally utilized to train an end-to-end explanation generator model. Extensive experiments are conducted to demonstrate the effectiveness of TAGExplainer in producing faithful and concise natural language explanations.

To address the semantic inconsistency issue with SAM or other single-image segmentation models handling image sequences, we introduce BYOCL. This novel model outperforms SAM in extensive experiments, showcasing its Hierarchical prototype capabilities across CLIP and other representations. BYOCL significantly reduces time and space consumption by dividing inputs into smaller batches, achieving exponential time reduction compared to previous methods. Our approach leverages the SAM image encoder for feature extraction, followed by Intra-Batch and Inter-Batch clustering algorithms. Extensive experiments demonstrate that BYOCL far exceeds the previous state-of-the-art single image segmentation model. Our work is the first to apply consistent segmentation using foundation models without requiring training, utilizing plug-and-play modules for any latent space, making our method highly efficientModels are available at \href{https://github.com/cyt1202/BYOCL.git, Comment: 5 pages, 5 figures

Vision Foundation Models (VFMs) have demonstrated outstanding performance on numerous downstream tasks. However, due to their inherent representation biases originating from different training paradigms, VFMs exhibit advantages and disadvantages across distinct vision tasks. Although amalgamating the strengths of multiple VFMs for downstream tasks is an intuitive strategy, effectively exploiting these biases remains a significant challenge. In this paper, we propose a novel and versatile "Swiss Army Knife" (SAK) solution, which adaptively distills knowledge from a committee of VFMs to enhance multi-task learning. Unlike existing methods that use a single backbone for knowledge transfer, our approach preserves the unique representation bias of each teacher by collaborating the lightweight Teacher-Specific Adapter Path modules with the Teacher-Agnostic Stem. Through dynamic selection and combination of representations with Mixture-of-Representations Routers, our SAK is capable of synergizing the complementary strengths of multiple VFMs. Extensive experiments show that our SAK remarkably outperforms prior state of the arts in multi-task learning by 10% on the NYUD-v2 benchmark, while also providing a flexible and robust framework that can readily accommodate more advanced model designs.

Large language models (LLMs) could be valuable personal AI agents across various domains, provided they can precisely follow user instructions. However, recent studies have shown significant limitations in LLMs' instruction-following capabilities, raising concerns about their reliability in high-stakes applications. Accurately estimating LLMs' uncertainty in adhering to instructions is critical to mitigating deployment risks. We present, to our knowledge, the first systematic evaluation of the uncertainty estimation abilities of LLMs in the context of instruction-following. Our study identifies key challenges with existing instruction-following benchmarks, where multiple factors are entangled with uncertainty stems from instruction-following, complicating the isolation and comparison across methods and models. To address these issues, we introduce a controlled evaluation setup with two benchmark versions of data, enabling a comprehensive comparison of uncertainty estimation methods under various conditions. Our findings show that existing uncertainty methods struggle, particularly when models make subtle errors in instruction following. While internal model states provide some improvement, they remain inadequate in more complex scenarios. The insights from our controlled evaluation setups provide a crucial understanding of LLMs' limitations and potential for uncertainty estimation in instruction-following tasks, paving the way for more trustworthy AI agents.

Synthetic data has been widely used to train large language models, but their generative nature inevitably introduces noisy, non-informative, and misleading learning signals. In this paper, we propose Montessori-Instruct, a novel data synthesis framework that tailors the data synthesis ability of the teacher language model toward the student language model's learning process. Specifically, we utilize local data influence of synthetic training data points on students to characterize students' learning preferences. Then, we train the teacher model with Direct Preference Optimization (DPO) to generate synthetic data tailored toward student learning preferences. Experiments with Llama3-8B-Instruct (teacher) and Llama3-8B (student) on Alpaca Eval and MT-Bench demonstrate that Montessori-Instruct significantly outperforms standard synthesis methods by 18.35\% and 46.24\% relatively. Our method also beats data synthesized by a stronger teacher model, GPT-4o. Further analysis confirms the benefits of teacher's learning to generate more influential training data in the student's improved learning, the advantages of local data influence in accurately measuring student preferences, and the robustness of Montessori-Instruct across different student models. Our code and data are open-sourced at https://github.com/cxcscmu/Montessori-Instruct., Comment: Codes and data are open-sourced at https://github.com/cxcscmu/Montessori-Instruct

Time series forecasting aids decision-making, especially for stakeholders who rely on accurate predictions, making it very important to understand and explain these models to ensure informed decisions. Traditional explainable AI (XAI) methods, which underline feature or temporal importance, often require expert knowledge. In contrast, natural language explanations (NLEs) are more accessible to laypeople. However, evaluating forecast NLEs is difficult due to the complex causal relationships in time series data. To address this, we introduce two new performance metrics based on simulatability, assessing how well a human surrogate can predict model forecasts using the explanations. Experiments show these metrics differentiate good from poor explanations and align with human judgments. Utilizing these metrics, we further evaluate the ability of state-of-the-art large language models (LLMs) to generate explanations for time series data, finding that numerical reasoning, rather than model size, is the main factor influencing explanation quality.

Dense embeddings deliver strong retrieval performance but often lack interpretability and controllability. This paper introduces a novel approach using sparse autoencoders (SAE) to interpret and control dense embeddings via the learned latent sparse features. Our key contribution is the development of a retrieval-oriented contrastive loss, which ensures the sparse latent features remain effective for retrieval tasks and thus meaningful to interpret. Experimental results demonstrate that both the learned latent sparse features and their reconstructed embeddings retain nearly the same retrieval accuracy as the original dense vectors, affirming their faithfulness. Our further examination of the sparse latent space reveals interesting features underlying the dense embeddings and we can control the retrieval behaviors via manipulating the latent sparse features, for example, prioritizing documents from specific perspectives in the retrieval results.

Text-rich visual understanding-the ability to process environments where dense textual content is integrated with visuals-is crucial for multimodal large language models (MLLMs) to interact effectively with structured environments. To enhance this capability, we propose synthesizing general multimodal instructions from webpage UIs using text-based large language models (LLMs). Despite lacking direct visual input, text-based LLMs are able to process structured text representations from webpage accessibility trees. These instructions are then paired with UI screenshots to train multimodal models. We introduce MultiUI, a dataset containing 7.3 million samples from 1 million websites, covering diverse multimodal tasks and UI layouts. Models trained on MultiUI not only excel in web UI tasks-achieving up to a 48% improvement on VisualWebBench and a 19.1% boost in element accuracy on a web agent dataset Mind2Web-but also generalize surprisingly well to non-web UI tasks and even to non-UI domains, such as document understanding, OCR, and chart interpretation. These results highlight the broad applicability of web UI data for advancing text-rich visual understanding across various scenarios.

Molecular optimization (MO) is a crucial stage in drug discovery in which task-oriented generated molecules are optimized to meet practical industrial requirements. Existing mainstream MO approaches primarily utilize external property predictors to guide iterative property optimization. However, learning all molecular samples in the vast chemical space is unrealistic for predictors. As a result, errors and noise are inevitably introduced during property prediction due to the nature of approximation. This leads to discrepancy accumulation, generalization reduction and suboptimal molecular candidates. In this paper, we propose a text-guided multi-property molecular optimization method utilizing transformer-based diffusion language model (TransDLM). TransDLM leverages standardized chemical nomenclature as semantic representations of molecules and implicitly embeds property requirements into textual descriptions, thereby preventing error propagation during diffusion process. Guided by physically and chemically detailed textual descriptions, TransDLM samples and optimizes encoded source molecules, retaining core scaffolds of source molecules and ensuring structural similarities. Moreover, TransDLM enables simultaneous sampling of multiple molecules, making it ideal for scalable, efficient large-scale optimization through distributed computation on web platforms. Furthermore, our approach surpasses state-of-the-art methods in optimizing molecular structural similarity and enhancing chemical properties on the benchmark dataset. The code is available at: https://anonymous.4open.science/r/TransDLM-A901.

Retrieval-Augmented Generation (RAG) has proven its effectiveness in mitigating hallucinations in Large Language Models (LLMs) by retrieving knowledge from external resources. To adapt LLMs for RAG pipelines, current approaches use instruction tuning to optimize LLMs, improving their ability to utilize retrieved knowledge. This supervised fine-tuning (SFT) approach focuses on equipping LLMs to handle diverse RAG tasks using different instructions. However, it trains RAG modules to overfit training signals and overlooks the varying data preferences among agents within the RAG system. In this paper, we propose a Differentiable Data Rewards (DDR) method, which end-to-end trains RAG systems by aligning data preferences between different RAG modules. DDR works by collecting the rewards to optimize each agent with a rollout method. This method prompts agents to sample some potential responses as perturbations, evaluates the impact of these perturbations on the whole RAG system, and subsequently optimizes the agent to produce outputs that improve the performance of the RAG system. Our experiments on various knowledge-intensive tasks demonstrate that DDR significantly outperforms the SFT method, particularly for LLMs with smaller-scale parameters that depend more on the retrieved knowledge. Additionally, DDR exhibits a stronger capability to align the data preference between RAG modules. The DDR method makes generation module more effective in extracting key information from documents and mitigating conflicts between parametric memory and external knowledge. All codes are available at https://github.com/OpenMatch/RAG-DDR.

Vision-Language Models (VLMs) often generate plausible but incorrect responses to visual queries. However, reliably quantifying the effect of such hallucinations in free-form responses to open-ended queries is challenging as it requires visually verifying each claim within the response. We propose Programmatic VLM Evaluation (PROVE), a new benchmarking paradigm for evaluating VLM responses to open-ended queries. To construct PROVE, we provide a large language model (LLM) with a high-fidelity scene-graph representation constructed from a hyper-detailed image caption, and prompt it to generate diverse question-answer (QA) pairs, as well as programs that can be executed over the scene graph object to verify each QA pair. We thus construct a benchmark of 10.5k challenging but visually grounded QA pairs. Next, to evaluate free-form model responses to queries in PROVE, we propose a programmatic evaluation strategy that measures both the helpfulness and truthfulness of a response within a unified scene graph-based framework. We benchmark the helpfulness-truthfulness trade-offs of a range of VLMs on PROVE, finding that very few are in-fact able to achieve a good balance between the two. Project page: \url{https://prove-explorer.netlify.app/}.

Molecular gas is the key probe for the complex interaction between the accretion of black holes and star formation of the host galaxy of active galactic nuclei (AGN). The molecular gas discovered around the AGN indicates that this gas is providing fuel for the AGN. According to the theoretical model of the relativistic jet, the spin of a black hole enhances the relativistic jet of AGN. The spin of the black hole is used as an indicator of AGN activity. Therefore, we study the relationship between the activity of AGN and molecular gas. We find a significant strong correlation between molecular gas fraction and CO luminosity and black hole spin for the early-type galaxies. However, there is no correlation between molecular gas fraction and CO luminosity and black hole spin for the late-type galaxies. These results indicate that the spin of black holes mainly regulates the accretion of molecular gas in massive early-type galaxies. The activity of AGN depends on the amount of gas., Comment: 8 pages, 4 figures, accept for publication in Monthly Notices of the Royal Astronomical Society

Anomaly detection in multivariate time series is challenging as heterogeneous subsequence anomalies may occur. Reconstruction-based methods, which focus on learning nomral patterns in the frequency domain to detect diverse abnormal subsequences, achieve promising resutls, while still falling short on capturing fine-grained frequency characteristics and channel correlations. To contend with the limitations, we introduce CATCH, a framework based on frequency patching. We propose to patchify the frequency domain into frequency bands, which enhances its ability to capture fine-grained frequency characteristics. To perceive appropriate channel correlations, we propose a Channel Fusion Module (CFM), which features a patch-wise mask generator and a masked-attention mechanism. Driven by a bi-level multi-objective optimization algorithm, the CFM is encouraged to iteratively discover appropriate patch-wise channel correlations, and to cluster relevant channels while isolating adverse effects from irrelevant channels. Extensive experiments on 9 real-world datasets and 12 synthetic datasets demonstrate that CATCH achieves state-of-the-art performance.

The space-air-ground integrated network (SAGIN) greatly improves coverage and reliability for millimeter-wave (mmWave) communication in high-speed railway (HSR) scenarios. However, a significant challenge arises in the transmission scheduling due to the rapid changes in channel state, link selection for train mobile relays (MRs), and order of the flow scheduling. To tackle this challenge, we introduce an optimization problem focused on maximizing the weighted sum completed flows that satisfy the quality of service (QoS) requirements for HSR mmWave communication in SAGIN. To facilitate the simultaneous scheduling of flows by base station-MR (BS-MR), satellite-airship-MR, and satellite-MR links, we propose a link selection algorithm, which can help each flow choose a suitable set of links in every frame and determine whether the BS networks need the assistance of the satellite and airship. Furthermore, taking into account the priority and occupied time slots (TSs) resource of different flows, we propose a multi-link weighted flow scheduling (MWFS) algorithm. This algorithm not only prioritizes scheduling high-priority flows but also aims to maximize the weighted sum completed flows for MRs. Our simulation results confirm that the proposed algorithm significantly increases the weighted sum completed flows and the total transmitted bits. Additionally, the proposed algorithm can achieve the optimal flow transmission in different link switching periods and enhance the scheduling of high-priority flows compared to other algorithms., Comment: 16 pages, 15 figures, IEEE Transactions on Vehicular Technology

As large language models (LLMs) become increasingly accessible in many countries, it is essential to align them to serve pluralistic human values across cultures. However, pluralistic culture alignment in LLMs remain an open problem. In this paper, we propose CultureSPA, a Self-Pluralising Culture Alignment framework that allows LLMs to simultaneously align to pluralistic cultures. The framework first generates questions on various culture topics, then yields LLM outputs in response to these generated questions under both culture-aware and culture-unaware settings. By comparing culture-aware/unaware outputs, we are able to detect and collect culture-related instances. These instances are employed to fine-tune LLMs to serve pluralistic cultures in either a culture-joint or culture-specific way. Extensive experiments demonstrate that CultureSPA significantly improves the alignment of LLMs to diverse cultures without compromising general abilities. And further improvements can be achieved if CultureSPA is combined with advanced prompt engineering techniques. Comparisons between culture-joint and culture-specific tuning strategies, along with variations in data quality and quantity, illustrate the robustness of our method. We also explore the mechanisms underlying CultureSPA and the relations between different cultures it reflects., Comment: Implementation for the paper: https://github.com/shaoyangxu/CultureSPA

Retrieval-Augmented Generation (RAG), while serving as a viable complement to large language models (LLMs), often overlooks the crucial aspect of text chunking within its pipeline, which impacts the quality of knowledge-intensive tasks. This paper introduces the concept of Meta-Chunking, which refers to a granularity between sentences and paragraphs, consisting of a collection of sentences within a paragraph that have deep linguistic logical connections. To implement Meta-Chunking, we designed two strategies based on LLMs: Margin Sampling Chunking and Perplexity Chunking. The former employs LLMs to perform binary classification on whether consecutive sentences need to be segmented, making decisions based on the probability difference obtained from margin sampling. The latter precisely identifies text chunk boundaries by analyzing the characteristics of perplexity distribution. Additionally, considering the inherent complexity of different texts, we propose a strategy that combines Meta-Chunking with dynamic merging to achieve a balance between fine-grained and coarse-grained text chunking. Experiments conducted on eleven datasets demonstrate that Meta-Chunking can more efficiently improve the performance of single-hop and multi-hop question answering based on RAG. For instance, on the 2WikiMultihopQA dataset, it outperforms similarity chunking by 1.32 while only consuming 45.8% of the time. Our code is available at https://github.com/IAAR-Shanghai/Meta-Chunking.

Vision Language Models (VLMs) often struggle with culture-specific knowledge, particularly in languages other than English and in underrepresented cultural contexts. To evaluate their understanding of such knowledge, we introduce WorldCuisines, a massive-scale benchmark for multilingual and multicultural, visually grounded language understanding. This benchmark includes a visual question answering (VQA) dataset with text-image pairs across 30 languages and dialects, spanning 9 language families and featuring over 1 million data points, making it the largest multicultural VQA benchmark to date. It includes tasks for identifying dish names and their origins. We provide evaluation datasets in two sizes (12k and 60k instances) alongside a training dataset (1 million instances). Our findings show that while VLMs perform better with correct location context, they struggle with adversarial contexts and predicting specific regional cuisines and languages. To support future research, we release a knowledge base with annotated food entries and images along with the VQA data., Comment: Preprint

Benefiting from the revolutionary advances in large language models (LLMs) and foundational vision models, large vision-language models (LVLMs) have also made significant progress. However, current benchmarks focus on tasks that evaluating only a single aspect of LVLM capabilities (e.g., recognition, detection, understanding). These tasks fail to fully demonstrate LVLMs' potential in complex application scenarios. To comprehensively assess the performance of existing LVLMs, we propose a more challenging task called the Flow Text with Image Insertion task (FTII). This task requires LVLMs to simultaneously possess outstanding abilities in image comprehension, instruction understanding, and long-text interpretation. Specifically, given several text paragraphs and a set of candidate images, as the text paragraphs accumulate, the LVLMs are required to select the most suitable image from the candidates to insert after the corresponding paragraph. Constructing a benchmark for such a task is highly challenging, particularly in determining the sequence of flowing text and images. To address this challenge, we turn to professional news reports, which naturally contain a gold standard for image-text sequences. Based on this, we introduce the Flow Text with Image Insertion Benchmark (FTII-Bench), which includes 318 high-quality Chinese image-text news articles and 307 high-quality English image-text news articles, covering 10 different news domains. Using these 625 high-quality articles, we construct problems of two different types with multiple levels of difficulty. Furthermore, we establish two different evaluation pipelines based on the CLIP model and existing LVLMs. We evaluate 9 open-source and 2 closed-source LVLMs as well as 2 CLIP-based models. Results indicate that even the most advanced models (e.g., GPT-4o) face significant challenges when tackling the FTII task., Comment: Work in progress. 9 pages, 3 figures