Search

Text-Video Retrieval (TVR) aims to align and associate relevant video content with corresponding natural language queries. Most existing TVR methods are based on large-scale pre-trained vision-language models (e.g., CLIP). However, due to the inherent plain structure of CLIP, few TVR methods explore the multi-scale representations which offer richer contextual information for a more thorough understanding. To this end, we propose MUSE, a multi-scale mamba with linear computational complexity for efficient cross-resolution modeling. Specifically, the multi-scale representations are generated by applying a feature pyramid on the last single-scale feature map. Then, we employ the Mamba structure as an efficient multi-scale learner to jointly learn scale-wise representations. Furthermore, we conduct comprehensive studies to investigate different model structures and designs. Extensive results on three popular benchmarks have validated the superiority of MUSE., Comment: 8 pages

$\pi$ modes are unique topological edge states appearing in Floquet systems with periodic modulations of the underlying lattice structure in evolution variable, such as dynamically modulated Su-Schrieffer-Heeger (SSH) lattices. These edge states are anomalous states usually appearing between Floquet replicas of the same band, even if standard topological index remains zero for this band. While linear and nonlinear $\pi$ modes were observed in conservative systems, they have never been studied in nonlinear regime in the non-Hermitian systems with structured gain and losses. Here we show that SSH waveguide array with periodically oscillating waveguide positions in propagation direction and with parity-time symmetric refractive index landscape, can support $\pi$ modes that are damped or amplified at different ends of the array. By including nonlinearity and nonlinear absorption into our continuous system, we achieve stable lasing in $\pi$ mode at one end of the array. The representative feature of this system is that lasing in it is thresholdless and it occurs even at low gain-loss amplitudes. The degree of localization of lasing $\pi$ modes can be flexibly controlled by the amplitude of transverse waveguide oscillations. This work therefore introduces a new type of topological Floquet laser and a route to manipulation of $\pi$ modes by structured gain and losses., Comment: 9 pages, 3 figures, to appear in APL Photonics

We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development.

Recent advances in large language models (LLMs) have increased the demand for comprehensive benchmarks to evaluate their capabilities as human-like agents. Existing benchmarks, while useful, often focus on specific application scenarios, emphasizing task completion but failing to dissect the underlying skills that drive these outcomes. This lack of granularity makes it difficult to deeply discern where failures stem from. Additionally, setting up these environments requires considerable effort, and issues of unreliability and reproducibility sometimes arise, especially in interactive tasks. To address these limitations, we introduce the Massive Multitask Agent Understanding (MMAU) benchmark, featuring comprehensive offline tasks that eliminate the need for complex environment setups. It evaluates models across five domains, including Tool-use, Directed Acyclic Graph (DAG) QA, Data Science and Machine Learning coding, Contest-level programming and Mathematics, and covers five essential capabilities: Understanding, Reasoning, Planning, Problem-solving, and Self-correction. With a total of 20 meticulously designed tasks encompassing over 3K distinct prompts, MMAU provides a comprehensive framework for evaluating the strengths and limitations of LLM agents. By testing 18 representative models on MMAU, we provide deep and insightful analyses. Ultimately, MMAU not only sheds light on the capabilities and limitations of LLM agents but also enhances the interpretability of their performance. Datasets and evaluation scripts of MMAU are released at https://github.com/apple/axlearn/tree/main/docs/research/mmau.

In recent years, Multi-View Clustering (MVC) has attracted increasing attention for its potential to reduce the annotation burden associated with large datasets. The aim of MVC is to exploit the inherent consistency and complementarity among different views, thereby integrating information from multiple perspectives to improve clustering outcomes. Despite extensive research in MVC, most existing methods focus predominantly on harnessing complementary information across views to enhance clustering effectiveness, often neglecting the structural information among samples, which is crucial for exploring sample correlations. To address this gap, we introduce a novel framework, termed Structured Latent Representation Learning based Multi-View Clustering method (SLRL). SLRL leverages both the complementary and structural information. Initially, it learns a common latent representation for all views. Subsequently, to exploit the structural information among samples, a k-nearest neighbor graph is constructed from this common latent representation. This graph facilitates enhanced sample interaction through graph learning techniques, leading to a structured latent representation optimized for clustering. Extensive experiments demonstrate that SLRL not only competes well with existing methods but also sets new benchmarks in various multi-view datasets.

The type-II Dirac cone is a special feature of the band structure, whose Fermi level is represented by a pair of crossing lines. It has been demonstrated that such a structure is useful for investigating topological edge solitons, and more specifically, for mimicking the Kline tunneling. However, it is still not clear what the interplay between type-II Dirac cones and the non-Hermiticity mechanism will result in. Here, this question is addressed; in particular, we report the $\mathcal{PT}$-symmetric photonic lattices with type-II Dirac cones for the first time. We identify a slope-exceptional ring and name it the type-II exceptional ring. We display the restoration of the $\mathcal{PT}$ symmetry of the lattice by reducing the separation between the sites in the unit cell. Curiously, the amplitude of the beam during propagation in the non-Hermitian lattice with $\mathcal{PT}$ symmetry only decays because of diffraction, whereas in the $\mathcal{PT}$ symmetry-broken lattice it will be amplified, even though the beam still diffracts. This work establishes the link between the non-Hermiticity mechanism and the violation of Lorentz invariance in these physical systems., Comment: 5 pages, 4 figures, to appear in Optics Letters. Comments are welcome

In category learning, a growing body of literature has increasingly focused on exploring the impacts of interleaving in contrast to blocking. The sequential attention hypothesis posits that interleaving draws attention to the differences between categories while blocking directs attention toward similarities within categories. Although a recent study underscores the joint influence of memory and attentional factors on sequencing effects, there remains a scarcity of effective computational models integrating both attentional and memory considerations to comprehensively understand the effect of training sequences on students' performance. This study introduces a novel integration of attentional factors and spacing into the logistic knowledge tracing (LKT) models to monitor students' performance across different training sequences (interleaving and blocking). Attentional factors were incorporated by recording the counts of comparisons between adjacent trials, considering whether they belong to the same or different category. Several features were employed to account for temporal spacing. We used cross-validations to test the model fit and predictions on the learning session and posttest. Our findings reveal that incorporating both attentional factors and spacing features in the Additive Factors Model (AFM) significantly enhances its capacity to capture the effects of interleaving and blocking and demonstrates superior predictive accuracy for students' learning outcomes. By bridging the gap between attentional factors and memory processes, our computational approach offers a more comprehensive framework for understanding and predicting category learning outcomes in educational settings., Comment: 7 pages, 3 figures, Educational Data Mining 2024

Real-time detection and prediction of extreme weather protect human lives and infrastructure. Traditional methods rely on numerical threshold setting and manual interpretation of weather heatmaps with Geographic Information Systems (GIS), which can be slow and error-prone. Our research redefines Extreme Weather Events Detection (EWED) by framing it as a Visual Question Answering (VQA) problem, thereby introducing a more precise and automated solution. Leveraging Vision-Language Models (VLM) to simultaneously process visual and textual data, we offer an effective aid to enhance the analysis process of weather heatmaps. Our initial assessment of general-purpose VLMs (e.g., GPT-4-Vision) on EWED revealed poor performance, characterized by low accuracy and frequent hallucinations due to inadequate color differentiation and insufficient meteorological knowledge. To address these challenges, we introduce ClimateIQA, the first meteorological VQA dataset, which includes 8,760 wind gust heatmaps and 254,040 question-answer pairs covering four question types, both generated from the latest climate reanalysis data. We also propose Sparse Position and Outline Tracking (SPOT), an innovative technique that leverages OpenCV and K-Means clustering to capture and depict color contours in heatmaps, providing ClimateIQA with more accurate color spatial location information. Finally, we present Climate-Zoo, the first meteorological VLM collection, which adapts VLMs to meteorological applications using the ClimateIQA dataset. Experiment results demonstrate that models from Climate-Zoo substantially outperform state-of-the-art general VLMs, achieving an accuracy increase from 0% to over 90% in EWED verification. The datasets and models in this study are publicly available for future climate science research: https://github.com/AlexJJJChen/Climate-Zoo.

Existing mainstream approaches follow the encoder-decoder paradigm for generating radiology reports. They focus on improving the network structure of encoders and decoders, which leads to two shortcomings: overlooking the modality gap and ignoring report content constraints. In this paper, we proposed Textual Inversion and Self-supervised Refinement (TISR) to address the above two issues. Specifically, textual inversion can project text and image into the same space by representing images as pseudo words to eliminate the cross-modeling gap. Subsequently, self-supervised refinement refines these pseudo words through contrastive loss computation between images and texts, enhancing the fidelity of generated reports to images. Notably, TISR is orthogonal to most existing methods, plug-and-play. We conduct experiments on two widely-used public datasets and achieve significant improvements on various baselines, which demonstrates the effectiveness and generalization of TISR. The code will be available soon., Comment: This paper has been early accepted by MICCAI 2024!

Sign language video retrieval plays a key role in facilitating information access for the deaf community. Despite significant advances in video-text retrieval, the complexity and inherent uncertainty of sign language preclude the direct application of these techniques. Previous methods achieve the mapping between sign language video and text through fine-grained modal alignment. However, due to the scarcity of fine-grained annotation, the uncertainty inherent in sign language video is underestimated, limiting the further development of sign language retrieval tasks. To address this challenge, we propose a novel Uncertainty-aware Probability Distribution Retrieval (UPRet), that conceptualizes the mapping process of sign language video and text in terms of probability distributions, explores their potential interrelationships, and enables flexible mappings. Experiments on three benchmarks demonstrate the effectiveness of our method, which achieves state-of-the-art results on How2Sign (59.1%), PHOENIX-2014T (72.0%), and CSL-Daily (78.4%).

Text-Video Retrieval (TVR) aims to align relevant video content with natural language queries. To date, most state-of-the-art TVR methods learn image-to-video transfer learning based on large-scale pre-trained visionlanguage models (e.g., CLIP). However, fully fine-tuning these pre-trained models for TVR incurs prohibitively expensive computation costs. To this end, we propose to conduct efficient text-video Retrieval with a sparse-andcorrelated AdaPter (RAP), i.e., fine-tuning the pre-trained model with a few parameterized layers. To accommodate the text-video scenario, we equip our RAP with two indispensable characteristics: temporal sparsity and correlation. Specifically, we propose a low-rank modulation module to refine the per-image features from the frozen CLIP backbone, which accentuates salient frames within the video features while alleviating temporal redundancy. Besides, we introduce an asynchronous self-attention mechanism that first selects the top responsive visual patches and augments the correlation modeling between them with learnable temporal and patch offsets. Extensive experiments on four TVR datasets demonstrate that RAP achieves superior or comparable performance compared to the fully fine-tuned counterpart and other parameter-efficient fine-tuning methods., Comment: Accepted by ACL 2024 Findings

Let $\mathrm{SLAut}(\mathbb{F}_{q}^{n})$ denote the group of all semilinear isometries on $\mathbb{F}_{q}^{n}$, where $q=p^{e}$ is a prime power. Matrix-product (MP) codes are a class of long classical codes generated by combining several commensurate classical codes with a defining matrix. We give an explicit formula for calculating the dimension of the $\sigma$ hull of a MP code. As a result, we give necessary and sufficient conditions for the MP codes to be $\sigma$ dual-containing and $\sigma$ self-orthogonal. We prove that $\mathrm{dim}_{\mathbb{F}_{q}}(\mathrm{Hull}_{\sigma}(\mathcal{C}))=\mathrm{dim}_{\mathbb{F}_{q}}(\mathrm{Hull}_{\sigma}(\mathcal{C}^{\bot_{\sigma}}))$. We prove that for any integer $h$ with $\mathrm{max}\{0,k_{1}-k_{2}\}\leq h\leq \mathrm{dim}_{\mathbb{F}_{q}}(\mathcal{C}_{1}\cap\mathcal{C}_{2}^{\bot_{\sigma}})$, there exists a linear code $\mathcal{C}_{2,h}$ monomially equivalent to $\mathcal{C}_{2}$ such that $\mathrm{dim}_{\mathbb{F}_{q}}(\mathcal{C}_{1}\cap\mathcal{C}_{2,h}^{\bot_{\sigma}})=h$, where $\mathcal{C}_{i}$ is an $[n,k_{i}]_{q}$ linear code for $i=1,2$. We show that given an $[n,k,d]_{q}$ linear code $\mathcal{C}$, there exists a monomially equivalent $[n,k,d]_{q}$ linear code $\mathcal{C}_{h}$, whose $\sigma$ dual code has minimum distance $d'$, such that there exist an $[[n,k-h,d;n-k-h]]_{q}$ EAQECC and an $[[n,n-k-h,d';k-h]]_{q}$ EAQECC for every integer $h$ with $0\leq h\leq \mathrm{dim}_{\mathbb{F}_{q}}(\mathrm{Hull}_{\sigma}(\mathcal{C}))$. Based on this result, we present a general construction method for deriving EAQECCs with flexible parameters from MP codes related to $\sigma$ hulls.

The matrix-product (MP) code $\mathcal{C}_{A,k}:=[\mathcal{C}_{1},\mathcal{C}_{2},\ldots,\mathcal{C}_{k}]\cdot A$ with a non-singular by column (NSC) matrix $A$ plays an important role in constructing good quantum error-correcting codes. In this paper, we study the MP code when the defining matrix $A$ satisfies the condition that $AA^{\dag}$ is $(D,\tau)$-monomial. We give an explicit formula for calculating the dimension of the Hermitian hull of a MP code. We provide the necessary and sufficient conditions that a MP code is Hermitian dual-containing (HDC), almost Hermitian dual-containing (AHDC), Hermitian self-orthogonal (HSO), almost Hermitian self-orthogonal (AHSO), and Hermitian LCD, respectively. We theoretically determine the number of all possible ways involving the relationships among the constituent codes to yield a MP code with these properties, respectively. We give alternative necessary and sufficient conditions for a MP code to be AHDC and AHSO, respectively, and show several cases where a MP code is not AHDC or AHSO. We provide the construction methods of HDC and AHDC MP codes, including those with optimal minimum distance lower bounds.

Recent developments underscore the potential of textual information in enhancing learning models for a deeper understanding of medical visual semantics. However, language-guided medical image segmentation still faces a challenging issue. Previous works employ implicit and ambiguous architectures to embed textual information. This leads to segmentation results that are inconsistent with the semantics represented by the language, sometimes even diverging significantly. To this end, we propose a novel cross-modal conditioned Reconstruction for Language-guided Medical Image Segmentation (RecLMIS) to explicitly capture cross-modal interactions, which assumes that well-aligned medical visual features and medical notes can effectively reconstruct each other. We introduce conditioned interaction to adaptively predict patches and words of interest. Subsequently, they are utilized as conditioning factors for mutual reconstruction to align with regions described in the medical notes. Extensive experiments demonstrate the superiority of our RecLMIS, surpassing LViT by 3.74% mIoU on the publicly available MosMedData+ dataset and achieving an average increase of 1.89% mIoU for cross-domain tests on our QATA-CoV19 dataset. Simultaneously, we achieve a relative reduction of 20.2% in parameter count and a 55.5% decrease in computational load. The code will be available at https://github.com/ShashankHuang/RecLMIS.

State-of-the-art language models (LMs) sometimes generate non-factual hallucinations that misalign with world knowledge. To explore the mechanistic causes of these hallucinations, we create diagnostic datasets with subject-relation queries and adapt interpretability methods to trace hallucinations through internal model representations. We discover two general and distinct mechanistic causes of hallucinations shared across LMs (Llama-2, Pythia, GPT-J): 1) knowledge enrichment hallucinations: insufficient subject attribute knowledge in lower layer MLPs, and 2) answer extraction hallucinations: failure to select the correct object attribute in upper layer attention heads. We also found these two internal mechanistic causes of hallucinations are reflected in external manifestations. Based on insights from our mechanistic analysis, we propose a novel hallucination mitigation method through targeted restoration of the LM's internal fact recall pipeline, demonstrating superior performance compared to baselines.

PyTorch \texttt{2.x} introduces a compiler designed to accelerate deep learning programs. However, for machine learning researchers, adapting to the PyTorch compiler to full potential can be challenging. The compiler operates at the Python bytecode level, making it appear as an opaque box. To address this, we introduce \texttt{depyf}, a tool designed to demystify the inner workings of the PyTorch compiler. \texttt{depyf} decompiles bytecode generated by PyTorch back into equivalent source code, and establishes connections between in-memory code objects and their on-disk source code counterparts. This feature enables users to step through the source code line by line using debuggers, thus enhancing their understanding of the underlying processes. Notably, \texttt{depyf} is non-intrusive and user-friendly, primarily relying on two convenient context managers for its core functionality. The project is \href{https://github.com/thuml/depyf}{ openly available} and is recognized as a \href{https://pytorch.org/ecosystem/}{PyTorch ecosystem project}., Comment: 16 pages, 2 figures

Intelligent Tutoring Systems (ITSs) have significantly enhanced adult literacy training, a key factor for societal participation, employment opportunities, and lifelong learning. Our study investigates the application of advanced AI models, including Large Language Models (LLMs) like GPT-4, for predicting learning performance in adult literacy programs in ITSs. This research is motivated by the potential of LLMs to predict learning performance based on its inherent reasoning and computational capabilities. By using reading comprehension datasets from the ITS, AutoTutor, we evaluate the predictive capabilities of GPT-4 versus traditional machine learning methods in predicting learning performance through five-fold cross-validation techniques. Our findings show that the GPT-4 presents the competitive predictive abilities with traditional machine learning methods such as Bayesian Knowledge Tracing, Performance Factor Analysis, Sparse Factor Analysis Lite (SPARFA-Lite), tensor factorization and eXtreme Gradient Boosting (XGBoost). While XGBoost (trained on local machine) outperforms GPT-4 in predictive accuracy, GPT-4-selected XGBoost and its subsequent tuning on the GPT-4 platform demonstrates superior performance compared to local machine execution. Moreover, our investigation into hyper-parameter tuning by GPT-4 versus grid-search suggests comparable performance, albeit with less stability in the automated approach, using XGBoost as the case study. Our study contributes to the field by highlighting the potential of integrating LLMs with traditional machine learning models to enhance predictive accuracy and personalize adult literacy education, setting a foundation for future research in applying LLMs within ITSs., Comment: 26TH International Conference on Human-Computer Interaction

Aligning large language models (LLMs) with human expectations without human-annotated preference data is an important problem. In this paper, we propose a method to evaluate the response preference by using the output probabilities of response pairs under contrastive prompt pairs, which could achieve better performance on LLaMA2-7B and LLaMA2-13B compared to RLAIF. Based on this, we propose an automatic alignment method, Direct Large Model Alignment (DLMA). First, we use contrastive prompt pairs to automatically generate preference data. Then, we continue to evaluate the generated preference data using contrastive prompt pairs and calculate a self-rewarding score. Finally, we use the DPO algorithm to effectively align LLMs by combining this self-rewarding score. In the experimental stage, our DLMA method could surpass the \texttt{RLHF} method without relying on human-annotated preference data., Comment: 24 pages, 5 pages

In this paper, we establish a dynamical quantum state tomography framework. Under this framework, it is feasible to obtain complete knowledge of any unknown state of a $d$-level system via only an arbitrary operator of certain types of IC-POVMs in dimension $d$. We show that under the time-dependent average channel, we can acquire a collection of projective operators that is informationally complete (IC) and thus obtain the corresponding IC-POVMs. We show that under certain condition, it is possible to obtain infinite families of projective operators that are IC, and obtain infinite families of corresponding IC-POVMs; otherwise, the Zauner's conjecture is incorrect. We also show how to simulate a SIC-POVM on any unknown quantum state by using the time-dependent average channel., Comment: 23 pages, 1 table

In social recommender systems, it is crucial that the recommendation models provide equitable visibility for different demographic groups, such as gender or race. Most existing research has addressed this problem by only studying individual static snapshots of networks that typically change over time. To address this gap, we study the evolution of recommendation fairness over time and its relation to dynamic network properties. We examine three real-world dynamic networks by evaluating the fairness of six recommendation algorithms and analyzing the association between fairness and network properties over time. We further study how interventions on network properties influence fairness by examining counterfactual scenarios with alternative evolution outcomes and differing network properties. Our results on empirical datasets suggest that recommendation fairness improves over time, regardless of the recommendation method. We also find that two network properties, minority ratio, and homophily ratio, exhibit stable correlations with fairness over time. Our counterfactual study further suggests that an extreme homophily ratio potentially contributes to unfair recommendations even with a balanced minority ratio. Our work provides insights into the evolution of fairness within dynamic networks in social science. We believe that our findings will help system operators and policymakers to better comprehend the implications of temporal changes and interventions targeting fairness in social networks.

Learning performance data (e.g., quiz scores and attempts) is significant for understanding learner engagement and knowledge mastery level. However, the learning performance data collected from Intelligent Tutoring Systems (ITSs) often suffers from sparsity, impacting the accuracy of learner modeling and knowledge assessments. To address this, we introduce the 3DG framework (3-Dimensional tensor for Densification and Generation), a novel approach combining tensor factorization with advanced generative models, including Generative Adversarial Network (GAN) and Generative Pre-trained Transformer (GPT), for enhanced data imputation and augmentation. The framework operates by first representing the data as a three-dimensional tensor, capturing dimensions of learners, questions, and attempts. It then densifies the data through tensor factorization and augments it using Generative AI models, tailored to individual learning patterns identified via clustering. Applied to data from an AutoTutor lesson by the Center for the Study of Adult Literacy (CSAL), the 3DG framework effectively generated scalable, personalized simulations of learning performance. Comparative analysis revealed GAN's superior reliability over GPT-4 in this context, underscoring its potential in addressing data sparsity challenges in ITSs and contributing to the advancement of personalized educational technology.

Reinforcement learning (RL) can align language models with non-differentiable reward signals, such as human preferences. However, a major challenge arises from the sparsity of these reward signals - typically, there is only a single reward for an entire output. This sparsity of rewards can lead to inefficient and unstable learning. To address this challenge, our paper introduces an novel framework that utilizes the critique capability of Large Language Models (LLMs) to produce intermediate-step rewards during RL training. Our method involves coupling a policy model with a critic language model, which is responsible for providing comprehensive feedback of each part of the output. This feedback is then translated into token or span-level rewards that can be used to guide the RL training process. We investigate this approach under two different settings: one where the policy model is smaller and is paired with a more powerful critic model, and another where a single language model fulfills both roles. We assess our approach on three text generation tasks: sentiment control, language model detoxification, and summarization. Experimental results show that incorporating artificial intrinsic rewards significantly improve both sample efficiency and the overall performance of the policy model, supported by both automatic and human evaluation.

AI and NLP publication venues have increasingly encouraged researchers to reflect on possible ethical considerations, adverse impacts, and other responsible AI issues their work might engender. However, for specific NLP tasks our understanding of how prevalent such issues are, or when and why these issues are likely to arise, remains limited. Focusing on text summarization -- a common NLP task largely overlooked by the responsible AI community -- we examine research and reporting practices in the current literature. We conduct a multi-round qualitative analysis of 333 summarization papers from the ACL Anthology published between 2020-2022. We focus on how, which, and when responsible AI issues are covered, which relevant stakeholders are considered, and mismatches between stated and realized research goals. We also discuss current evaluation practices and consider how authors discuss the limitations of both prior work and their own work. Overall, we find that relatively few papers engage with possible stakeholders or contexts of use, which limits their consideration of potential downstream adverse impacts or other responsible AI issues. Based on our findings, we make recommendations on concrete practices and research directions.

Large Multimodal Models (LMMs) have demonstrated impressive performance across various vision and language tasks, yet their potential applications in recommendation tasks with visual assistance remain unexplored. To bridge this gap, we present a preliminary case study investigating the recommendation capabilities of GPT-4V(ison), a recently released LMM by OpenAI. We construct a series of qualitative test samples spanning multiple domains and employ these samples to assess the quality of GPT-4V's responses within recommendation scenarios. Evaluation results on these test samples prove that GPT-4V has remarkable zero-shot recommendation abilities across diverse domains, thanks to its robust visual-text comprehension capabilities and extensive general knowledge. However, we have also identified some limitations in using GPT-4V for recommendations, including a tendency to provide similar responses when given similar inputs. This report concludes with an in-depth discussion of the challenges and research opportunities associated with utilizing GPT-4V in recommendation scenarios. Our objective is to explore the potential of extending LMMs from vision and language tasks to recommendation tasks. We hope to inspire further research into next-generation multimodal generative recommendation models, which can enhance user experiences by offering greater diversity and interactivity. All images and prompts used in this report will be accessible at https://github.com/PALIN2018/Evaluate_GPT-4V_Rec., Comment: In Progress

While large language models (LLMs) have achieved impressive performance in generating fluent and realistic text, controlling the generated text so that it exhibits properties such as safety, factuality, and non-toxicity remains challenging. % such as DExperts, GeDi, and rectification Existing decoding-based methods are static in terms of the dimension of control; if the target subject is changed, they require new training. Moreover, it can quickly become prohibitive to concurrently control multiple subjects. In this work, we introduce SF-GEN, which is grounded in two primary concepts: successor features (SFs) to decouple the LLM's dynamics from task-specific rewards, and language model rectification to proportionally adjust the probability of selecting a token based on the likelihood that the finished text becomes undesired. SF-GEN seamlessly integrates the two to enable dynamic steering of text generation with no need to alter the LLM's parameters. Thanks to the decoupling effect induced by successor features, our method proves to be memory-wise and computationally efficient for training as well as decoding, especially when dealing with multiple target subjects. To the best of our knowledge, our research represents the first application of successor features in text generation. In addition to its computational efficiency, the resultant language produced by our method is comparable to the SOTA (and outperforms baselines) in both control measures as well as language quality, which we demonstrate through a series of experiments in various controllable text generation tasks.

Video Referring Expression Comprehension (REC) aims to localize a target object in videos based on the queried natural language. Recent improvements in video REC have been made using Transformer-based methods with learnable queries. However, we contend that this naive query design is not ideal given the open-world nature of video REC brought by text supervision. With numerous potential semantic categories, relying on only a few slow-updated queries is insufficient to characterize them. Our solution to this problem is to create dynamic queries that are conditioned on both the input video and language to model the diverse objects referred to. Specifically, we place a fixed number of learnable bounding boxes throughout the frame and use corresponding region features to provide prior information. Also, we noticed that current query features overlook the importance of cross-modal alignment. To address this, we align specific phrases in the sentence with semantically relevant visual areas, annotating them in existing video datasets (VID-Sentence and VidSTG). By incorporating these two designs, our proposed model (called ConFormer) outperforms other models on widely benchmarked datasets. For example, in the testing split of VID-Sentence dataset, ConFormer achieves 8.75% absolute improvement on Accu.@0.6 compared to the previous state-of-the-art model., Comment: Accepted to ACM International Conference on Multimedia Workshop (ACM MM), 2023. arXiv admin note: substantial text overlap with arXiv:2210.02953

Integrating large language models (LLMs) into healthcare holds great potential but faces challenges. Pre-training LLMs from scratch for domains like medicine is resource-heavy and often unfeasible. On the other hand, sole reliance on Supervised Fine-tuning (SFT) can result in overconfident predictions and may not tap into domain-specific insights. In response, we present a multi-stage training method combining Domain-specific Continued Pre-training (DCPT), SFT, and Direct Preference Optimization (DPO). In addition, we publish a 3Gb Chinese Medicine (ChiMed) dataset, encompassing medical question answering, plain texts, knowledge graphs, and dialogues, segmented into three training stages. The medical LLM trained with our pipeline, Qilin-Med, shows substantial performance improvement. In the CPT and SFT phases, Qilin-Med achieved 38.4% and 40.0% accuracy on the CMExam test set, respectively. It outperformed the basemodel Baichuan-7B (accuracy: 33.5%), by 7.5%. In the DPO phase, it scored 16.66 in BLEU-1 and 27.44 in ROUGE-1 on the Huatuo-26M test set, bringing further improvement to the SFT phase (12.69 in BLEU-1 and 24.21 in ROUGE-1). Additionally, we have further enhanced the model's performance through the Retrieval Augmented Generation (RAG) approach. Experiments demonstrate that Qilin-Med-RAG achieves an accuracy rate of 42.8% on CMExam. These results highlight the contribution of our novel training approach in building LLMs for medical applications.

Large-scale web-crawled datasets are fundamental for the success of pre-training vision-language models, such as CLIP. However, the inherent noise and potential irrelevance of web-crawled AltTexts pose challenges in achieving precise image-text alignment. Existing methods utilizing large language models (LLMs) for caption rewriting have shown promise on small, curated datasets like CC3M and CC12M. This study introduces a scalable pipeline for noisy caption rewriting. Unlike recent LLM rewriting techniques, we emphasize the incorporation of visual concepts into captions, termed as Visual-enriched Captions (VeCap). To ensure data diversity, we propose a novel mixed training scheme that optimizes the utilization of AltTexts alongside newly generated VeCap. We showcase the adaptation of this method for training CLIP on large-scale web-crawled datasets, termed VeCLIP. Employing this cost-effective pipeline, we effortlessly scale our dataset up to 300 million samples named VeCap dataset. Our results show significant advantages in image-text alignment and overall model performance. For example, VeCLIP achieves up to +25.2% gain in COCO and Flickr30k retrieval tasks under the 12M setting. For data efficiency, VeCLIP achieves +3% gain while only using 14% of the data employed in the vanilla CLIP and 11% in ALIGN. We also note the VeCap data is complementary with other well curated datasets good for zero-shot classification tasks. When combining VeCap and DFN, our model can achieve strong performance on both of image-text retrieval and zero-shot classification tasks, e.g. 83.1% accuracy@1 on ImageNet zero-shot for a H/14 model. We release the pre-trained models at https://github.com/apple/ml-veclip., Comment: CV/ML

Symmetric informationally complete positive operator-valued measures (SIC-POVMs) in finite dimension $d$ are a particularly attractive case of informationally complete POVMs (IC-POVMs), which consist of $d^{2}$ subnormalized projectors with equal pairwise fidelity. However, it is difficult to construct SIC-POVMs, and it is not even clear whether there exists an infinite family of SIC-POVMs. To realize some possible applications in quantum information processing, Klappenecker et al. [37] introduced an approximate version of SIC-POVMs called approximately symmetric informationally complete POVMs (ASIC-POVMs). In this paper, we construct a class of $\varepsilon_{d}$-ASIC-POVMs in dimension $d=q$ and a class of $\varepsilon_{d}$-ASIC-POVMs in dimension $d=q+1$, respectively, where $q$ is a prime power. We prove that all $2$-to-$1$ perfect nonlinear (PN) functions can be used for constructing $\varepsilon_{q}$-ASIC-POVMs. We show that the set of vectors corresponding to the $\varepsilon_{q}$-ASIC-POVM forms a biangular frame. The construction of $\varepsilon_{q+1}$-ASIC-POVMs is based on a multiplicative character sum estimate called the Li bound. We show that the set of vectors corresponding to the $\varepsilon_{q+1}$-ASIC-POVM forms an asymptotically optimal codebook. We characterize "how close" the $\varepsilon_{q}$-ASIC-POVMs (resp. $\varepsilon_{q+1}$-ASIC-POVMs) are from being SIC-POVMs of dimension $q$ (resp. dimension $q+1$). Finally, we explain the significance of constructing $\varepsilon_{d}$-ASIC-POVMs.

The recent video grounding works attempt to introduce vanilla contrastive learning into video grounding. However, we claim that this naive solution is suboptimal. Contrastive learning requires two key properties: (1) \emph{alignment} of features of similar samples, and (2) \emph{uniformity} of the induced distribution of the normalized features on the hypersphere. Due to two annoying issues in video grounding: (1) the co-existence of some visual entities in both ground truth and other moments, \ie semantic overlapping; (2) only a few moments in the video are annotated, \ie sparse annotation dilemma, vanilla contrastive learning is unable to model the correlations between temporally distant moments and learned inconsistent video representations. Both characteristics lead to vanilla contrastive learning being unsuitable for video grounding. In this paper, we introduce Geodesic and Game Localization (G2L), a semantically aligned and uniform video grounding framework via geodesic and game theory. We quantify the correlations among moments leveraging the geodesic distance that guides the model to learn the correct cross-modal representations. Furthermore, from the novel perspective of game theory, we propose semantic Shapley interaction based on geodesic distance sampling to learn fine-grained semantic alignment in similar moments. Experiments on three benchmarks demonstrate the effectiveness of our method., Comment: ICCV2023 oral, release the code

Retrieval augmentation enables large language models to take advantage of external knowledge, for example on tasks like question answering and data imputation. However, the performance of such retrieval-augmented models is limited by the data quality of their underlying retrieval corpus. In this paper, we propose an algorithm based on multilinear extension for evaluating the data importance of retrieved data points. There are exponentially many terms in the multilinear extension, and one key contribution of this paper is a polynomial time algorithm that computes exactly, given a retrieval-augmented model with an additive utility function and a validation set, the data importance of data points in the retrieval corpus using the multilinear extension of the model's utility function. We further proposed an even more efficient ({\epsilon}, {\delta})-approximation algorithm. Our experimental results illustrate that we can enhance the performance of large language models by only pruning or reweighting the retrieval corpus, without requiring further training. For some tasks, this even allows a small model (e.g., GPT-JT), augmented with a search engine API, to outperform GPT-3.5 (without retrieval augmentation). Moreover, we show that weights based on multilinear extension can be computed efficiently in practice (e.g., in less than ten minutes for a corpus with 100 million elements).

Distinctive Image Captioning (DIC) -- generating distinctive captions that describe the unique details of a target image -- has received considerable attention over the last few years. A recent DIC method proposes to generate distinctive captions by comparing the target image with a set of semantic-similar reference images, i.e., reference-based DIC (Ref-DIC). It aims to force the generated captions to distinguish between the target image and the reference image. To ensure Ref-DIC models really perceive the unique objects (or attributes) in target images, we propose two new Ref-DIC benchmarks and develop a Transformer-based Ref-DIC baseline TransDIC. The model only extracts visual features from the target image, but also encodes the differences between objects in the target and reference images. Taking one step further, we propose a stronger TransDIC++, which consists of an extra contrastive learning module to make full use of the reference images. This new module is model-agnostic, which can be easily incorporated into various Ref-DIC architectures. Finally, for more trustworthy benchmarking, we propose a new evaluation metric named DisCIDEr for Ref-DIC, which evaluates both the accuracy and distinctiveness of the generated captions. Experimental results demonstrate that our TransDIC++ can generate distinctive captions. Besides, it outperforms several state-of-the-art models on the two new benchmarks over different metrics., Comment: arXiv admin note: substantial text overlap with arXiv:2207.11118

Despite progress in vision-based inspection algorithms, real-world industrial challenges -- specifically in data availability, quality, and complex production requirements -- often remain under-addressed. We introduce the VISION Datasets, a diverse collection of 14 industrial inspection datasets, uniquely poised to meet these challenges. Unlike previous datasets, VISION brings versatility to defect detection, offering annotation masks across all splits and catering to various detection methodologies. Our datasets also feature instance-segmentation annotation, enabling precise defect identification. With a total of 18k images encompassing 44 defect types, VISION strives to mirror a wide range of real-world production scenarios. By supporting two ongoing challenge competitions on the VISION Datasets, we hope to foster further advancements in vision-based industrial inspection.