Text-to-video (T2V) generation models have made significant progress in creating visually appealing videos. However, they struggle with generating coherent sequential narratives that require logical progression through multiple events. Existing T2V benchmarks primarily focus on visual quality metrics but fail to evaluate narrative coherence over extended sequences. To bridge this gap, we present SeqBench, a comprehensive benchmark for evaluating sequential narrative coherence in T2V generation. SeqBench includes a carefully designed dataset of 320 prompts spanning various narrative complexities, with 2,560 human-annotated videos generated from 8 state-of-the-art T2V models. Additionally, we design a Dynamic Temporal Graphs (DTG)-based automatic evaluation metric, which can efficiently capture long-range dependencies and temporal ordering while maintaining computational efficiency. Our DTG-based metric demonstrates a strong correlation with human annotations. Through systematic evaluation using SeqBench, we reveal critical limitations in current T2V models: failure to maintain consistent object states across multi-action sequences, physically implausible results in multi-object scenarios, and difficulties in preserving realistic timing and ordering relationships between sequential actions. SeqBench provides the first systematic framework for evaluating narrative coherence in T2V generation and offers concrete insights for improving sequential reasoning capabilities in future models. Please refer to https://videobench.github.io/SeqBench.github.io/ for more details.

Multi-modal multi-task learning holds significant promise in tackling complex diagnostic tasks and many significant medical imaging problems. It fulfills the needs in real-world diagnosis protocol to leverage information from different data sources and simultaneously perform mutually informative tasks. However, medical imaging domains introduce two key challenges: dynamic modality fusion and modality-task dependence. The quality and amount of task-related information from different modalities could vary significantly across patient samples, due to biological and demographic factors. Traditional fusion methods apply fixed combination strategies that fail to capture this dynamic relationship, potentially underutilizing modalities that carry stronger diagnostic signals for specific patients. Additionally, different clinical tasks may require dynamic feature selection and combination from various modalities, a phenomenon we term “modality-task dependence.” To address these issues, we propose M4oE, a novel Multi-modal Multi-task Mixture of Experts framework for precise Medical diagnosis. M4oE comprises Modality-Specific (MSoE) modules and a Modality-shared Modality-Task MoE (MToE) module. With collaboration from both modules, our model dynamically decomposes and learns distinct and shared information from different modalities and achieves dynamic fusion. MToE provides a joint probability model of modalities and tasks by using experts as a link and encourages experts to learn modality-task dependence via conditional mutual information loss. By doing so, M4oE offers sample and population-level interpretability of modality contributions. We evaluate M4oE on four public multi-modal medical benchmark datasets for solving two important medical diagnostic problems including breast cancer screening and retinal disease diagnosis. Results demonstrate our method's superiority over state-of-the-art methods under different metrics of classification and segmentation tasks like Accuracy, AUROC, AUPRC, and DICE.

Post-training quantization (PTQ) has emerged as a promising technique to reduce the cost of large language models (LLMs). Specifically, PTQ can effectively mitigate memory consumption and reduce computational overhead in LLMs. To meet the requirements of both high efficiency and performance across diverse scenarios, a comprehensive evaluation of quantized LLMs is essential to guide the selection of quantization methods. This paper presents a thorough evaluation of these factors by evaluating the effect of PTQ on Weight, Activation, and KV Cache on 11 model families, including OPT, LLaMA2, Falcon, Bloomz, Mistral, ChatGLM, Vicuna, LongChat, StableLM, Gemma, and Mamba, with parameters ranging from 125M to 180B. The evaluation encompasses five types of tasks: basic NLP, emergent ability, trustworthiness, dialogue, and long-context tasks. Moreover, we also evaluate the state-of-the-art (SOTA) quantization methods to demonstrate their applicability. Based on the extensive experiments, we systematically summarize the effect of quantization, provide recommendations to apply quantization techniques, and point out future directions. The code can be found in https://github.com/thu-nics/qllm-eval.

Whole slide image (WSI) classification is one of the important fields of digital pathology, and is generally solved as a weakly supervised learning problem by adopting multiple instance learning (MIL). However, a common but crucial challenge faced by existing MIL models is their inability to provide convincing explanations that can win the trust of pathologists and be applied to clinical diagnosis. In addition, most attention-based MIL models use attention scores to represent the importance of each patch in the WSI rather than inferring patch probabilities directly, which does not accurately detect the critical patches. To address these two challenges, we propose a ProtoTree based MIL model for WSI classification, called ProtoTree-MIL, where ProtoTree is an interpretable model that combines the advantages of prototype-learning and decision tree. ProtoTree-MIL not only explains why some patches are important for the final prediction through prototype-learning, but also provides global and local explanation through decision tree. We also propose a method to infer patch probabilities and measure their importance under the framework of ProtoTree-MIL. By conducting various experiments on three public WSI datasets, Camelyon16, TCGA-NSCLC, and TCGA-RCC, we demonstrate that our proposed ProtoTree-MIL can achieve a competitive performance to the state-of-the-art MIL models but provide more persuasive explanations than them. Explicitly generating patch probabilities also makes ProtoTree-MIL more accurate to detect the key patches than other attention-based MIL models. Specially, by evaluating our model on a real clinical gastritis and gastric cancer dataset, we show the explanations provided by ProtoTree-MIL are significant and faithful.

Domain generalization (DG) is about training models that generalize well under domain shift. Previous research on DG has been conducted mostly in single-source or multi-source settings. In this paper, we consider a third lesser-known setting where a training domain is endowed with a collection of pairs of examples that share the same semantic information. Such semantic sharing (SS) pairs can be created via data augmentation and then utilized for consistency regularization (CR). We present a theory showing CR is conducive to DG and propose a novel CR method called Logit Attribution Matching (LAM). We conduct experiments on five DG benchmarks and four pretrained models with SS pairs created by both generic and targeted data augmentation methods. LAM outperforms representative single/multi-source DG methods and various CR methods that leverage SS pairs. The code and data of this project are available at https://github.com/Gaohan123/LAM.

Machine learning systems produce biased results towards certain demographic groups, known as the fairness problem. Recent approaches to tackle this problem learn a latent code (i.e., representation) through disentangled representation learning and then discard the latent code dimensions correlated with sensitive attributes (e.g., gender). Nevertheless, these approaches may suffer from incomplete disentanglement and overlook proxy attributes (proxies for sensitive attributes) when processing real-world data, especially for unstructured data, causing performance degradation in fairness and loss of useful information for downstream tasks. In this paper, we propose a novel fairness framework that performs debiasing with regard to both sensitive attributes and proxy attributes, which boosts the prediction performance of downstream task models without complete disentanglement. The main idea is to, first, leverage gradient-based explanation to find two model focuses, 1) one focus for predicting sensitive attributes and 2) the other focus for predicting downstream task labels, and second, use them to perturb the latent code that guides the training of downstream task models towards fairness and utility goals. We show empirically that our framework works with both disentangled and non-disentangled representation learning methods and achieves better fairness-accuracy trade-off on unstructured and structured datasets than previous state-of-the-art approaches.

Network performance tomography uses a small number of strategically deployed monitors to infer the link performance in a large network. With the limited number of monitors, however, people usually can only estimate the bound rather than the exact values of network link performance. We aim at developing an effective solution to minimize the maximum error bound ($\mathcal{M}\mathcal{E}\mathcal{B}$) over all the links in the network. To achieve this, we develop a method that theoretically guarantees (1) the minimum number of monitors required to bring down the $\mathcal{M}\mathcal{E}\mathcal{B}$ over all unidentifiable links, and (2) the best places where these new monitors should be deployed. Using this method repeatedly, we can push down the $\mathcal{M}\mathcal{E}\mathcal{B}$ gradually until the desired level is reached. In addition, we develop a new sequential measurement technique that reduces the number of measurement paths and in the meantime guarantees the tightest link error bound. With extensive simulation over real-world network topology, we demonstrate the effectiveness and robustness of our solution in reducing the maximum link error bound with network performance tomography.

Deep learning has shown powerful performances in many fields, however its black-box nature hinders its further applications. In response, explainable artificial intelligence emerges, aiming to explain the predictions and behaviors of deep learning models. Among many explanation methods, counterfactual explanation has been identified as one of the best methods due to its resemblance to human cognitive process: to deliver an explanation by constructing a contrastive situation so that human may interpret the underlying mechanism by cognitively demonstrating the difference.

Network performance tomography infers performance metrics on internal network links with end-to-end measurements. Existing results in this domain are mainly Boolean-based, i.e., they check whether or not a link is identifiable, and return the exact value on identifiable links. If a link is not identifiable, the Boolean-based solution gives no performance result for the link. In this paper, we extend Boolean-based network tomography to bound-based network tomography where the lower and upper bounds are derived for unidentifiable links. We develop an efficient algorithm to obtain the tightest total error bound, and present a solution that can significantly reduce the total number of measurement paths required for deriving the tightest total error bound. Furthermore, we propose a method to deploy a new monitor such that the total error bound could be maximally reduced. Compared to the random monitor deployment and the monitor deployment that maximizes the total number of identifiable links, our monitor deployment method can lead to up to 15 and 2.4 times more reduction on total error bound, respectively.

Network performance tomography infers performance metrics on internal network links with end-to-end measurements. Existing results in this domain are mainly Boolean-based, i.e., they check whether or not a link is identifiable, and return the exact value on identifiable links. If a link is not identifiable, Boolean-based solution gives no performance result for the link. In this paper, we extend Boolean-based network tomography to bound-based network tomography where the lower and upper bounds are derived for unidentifiable links. We develop an efficient algorithm to obtain the tightest total error bound, and present a solution that can significantly reduce the total number of measurement paths required for deriving the tightest total error bound. Furthermore, we propose a method to deploy a new monitor over existing ones such that the total error bound could be maximally reduced. Compared to the random monitor deployment and the monitor deployment that maximizes the total number of identifiable links, our monitor deployment method can lead to up to 15 and 2.4 times more reduction on total error bound, respectively.

In the “network-as-a-service” paradigm, network operators have a strong need to know the metrics of critical paths running services to their users/tenants. However, it is usually prohibitive to directly measure the metrics of all such paths due to the measuring overhead. A practical solution is to use network tomography to infer the metrics of such paths based on observations from a small number of monitoring nodes. This problem is termed as path identifiability problem, a new problem that largely differs from existing link identifiability problems. we show that the new problem is harder than link identifiability problems, in the sense that fewer monitors are required for identifying the metrics of given paths than for identifying the metrics of links along the paths. To solve the problem, we develop sufficient and necessary conditions for the identifiability of a given set of interested paths, and design an efficient algorithm that deploys the minimum number of monitors. Experiments show a saving of up to 40% fewer monitors that guarantee the identifiability of a given set of paths.

The business value of information technology (IT) has been one of the top concerns of both practitioners and scholars for decades. Numerous studies have documented the positive effects of IT capability on organizational performance but our knowledge of the processes through which such gains are achieved remains limited due to a lack of focus on the business environment. Such a linkage therefore remains the subject of debate in the information systems literature. In this study, we fill this gap by investigating the mediating role of business process agility and the moderating roles of environmental factors. On the basis of matched survey data obtained from 214 IT and business executives from manufacturing firms in China, our analyses show that even though firm-wide IT capability presents the characteristics of rarity, appropriability, non-reproducibility, and non-substitutability, its impact on organizational performance is fully mediated by business process agility. Our results also show that the impact of the environment is multifaceted and nuanced. In particular, environmental hostility weakens the effect of IT capability on business process agility, while environmental complexity strengthens it. The theoretical and practical implications of this study, and its limitations, are also discussed.

MOTIVATION: Liquid chromatography-mass spectrometry-based metabolomics has gained importance in the life sciences, yet it is not supported by software tools for high throughput identification of metabolites based on their fragmentation spectra. An algorithm (ISIS: in silico identification software) and its implementation are presented and show great promise in generating in silico spectra of lipids for the purpose of structural identification. Instead of using chemical reaction rate equations or rules-based fragmentation libraries, the algorithm uses machine learning to find accurate bond cleavage rates in a mass spectrometer employing collision-induced dissociation tandem mass spectrometry. RESULTS: A preliminary test of the algorithm with 45 lipids from a subset of lipid classes shows both high sensitivity and specificity.