Visual generative models have achieved remarkable progress in synthesizing photorealistic images and videos, yet aligning their outputs with human preferences across critical dimensions remains a persistent challenge. Though reinforcement learning from human feedback offers promise for preference alignment, existing reward models for visual generation face limitations, including black-box scoring without interpretability and potentially resultant unexpected biases. We present VisionReward, a general framework for learning human visual preferences in both image and video generation. Specifically, we employ a hierarchical visual assessment framework to capture fine-grained human preferences, and leverages linear weighting to enable interpretable preference learning. Furthermore, we propose a multi-dimensional consistent strategy when using VisionReward as a reward model during preference optimization for visual generation. Experiments show that VisionReward can significantly outperform existing image and video reward models on both machine metrics and human evaluation. Notably, VisionReward surpasses VideoScore by 17.2% in preference prediction accuracy, and text-to-video models with VisionReward achieve a 31.6% higher pairwise win rate compared to the same models using VideoScore.

In recent years, vision language models (VLMs) have made significant advancements in video understanding. However, a crucial capability — fine-grained motion comprehension — remains under-explored in current benchmarks. To address this gap, we propose MotionBench, a comprehensive evaluation benchmark designed to assess the fine-grained motion comprehension of video understanding models. MotionBench evaluates models’ motion-level perception through six primary categories of motion-oriented question types and includes data collected from diverse sources, ensuring a broad representation of real-world video content. Experimental results reveal that existing VLMs perform poorly in understanding fine-grained motions. To enhance VLM’s ability to perceive fine-grained motion within a limited sequence length of LLM, we conduct extensive experiments reviewing VLM architectures optimized for video feature compression and propose a novel and efficient Through-Encoder (TE) Fusion method. Experiments show that higher frame rate inputs and TE Fusion yield improvements in motion understanding, yet there is still substantial room for enhancement. Our benchmark aims to guide and motivate the development of more capable video understanding models, emphasizing the importance of fine-grained motion comprehension. Project page: https://motion-bench.github.io.

Video generation models have achieved remarkable progress in text-to-video tasks. These models are typically trained on text-video pairs with highly detailed and carefully crafted descriptions, while real-world user inputs during inference are often concise, vague, or poorly structured. This gap makes prompt optimization crucial for generating high-quality videos. Current methods often rely on large language models (LLMs) to refine prompts through in-context learning, but suffer from several limitations: they may distort user intent, omit critical details, or introduce safety risks. Moreover, they optimize prompts without considering the impact on the final video quality, which can lead to suboptimal results. To address these issues, we introduce VPO, a principled framework that optimizes prompts based on three core principles: harmlessness, accuracy, and helpfulness. The generated prompts faithfully preserve user intents and, more importantly, enhance the safety and quality of generated videos. To achieve this, VPO employs a two-stage optimization approach. First, we construct and refine a supervised fine-tuning (SFT) dataset based on principles of safety and alignment. Second, we introduce both text-level and video-level feedback to further optimize the SFT model with preference learning. Our extensive experiments demonstrate that VPO significantly improves safety, alignment, and video quality compared to baseline methods. Moreover, VPO shows strong generalization across video generation models. Furthermore, we demonstrate that VPO could outperform and be combined with RLHF methods on video generation models, underscoring the effectiveness of VPO in aligning video generation models. Our code and data are publicly available at https://github.com/thu-coai/VPO.

We present CogVideoX, a large-scale text-to-video generation model based on diffusion transformer, which can generate 10-second continuous videos that align seamlessly with text prompts, with a frame rate of 16 fps and resolution of 768 x 1360 pixels. Previous video generation models often struggled with limited motion and short durations. It is especially difficult to generate videos with coherent narratives based on text. We propose several designs to address these issues. First, we introduce a 3D Variational Autoencoder (VAE) to compress videos across spatial and temporal dimensions, enhancing both the compression rate and video fidelity. Second, to improve text-video alignment, we propose an expert transformer with expert adaptive LayerNorm to facilitate the deep fusion between the two modalities. Third, by employing progressive training and multi-resolution frame packing, CogVideoX excels at generating coherent, long-duration videos with diverse shapes and dynamic movements. In addition, we develop an effective pipeline that includes various pre-processing strategies for text and video data. Our innovative video captioning model significantly improves generation quality and semantic alignment. Results show that CogVideoX achieves state-of-the-art performance in both automated benchmarks and human evaluation. We publish the code and model checkpoints of CogVideoX along with our VAE model and video captioning model at https://github.com/THUDM/CogVideo.

Log anomaly detection refers to the task that distinguishes the anomalous log messages from normal log messages. Transformer-based large language models (LLMs) are becoming popular for log anomaly detection because of their superb ability to understand complex and long language patterns. In this paper, we propose LogLLaMA, a novel framework that leverages LLaMA2. LogLLaMA is first finetuned on normal log messages from three large-scale datasets to learn their patterns. After finetuning, the model is capable of generating successive log messages given previous log messages. Our generative model is further trained to identify anomalous log messages using reinforcement learning (RL). The experimental results show that LogLLaMA outperforms the state-of-the-art approaches for anomaly detection on BGL, Thunderbird, and HDFS datasets.

Longitudinal medical image processing is a significant task to understand the dynamic changes of disease by taking and comparing image series over time, providing insights into how conditions evolve and enabling more accurate di-agnosis and treatment planning. While recent advance-ments in biomedical Vision-Language Pre-training (VLP) have enabled label-efficient representation learning with paired medical images and reports, existing methods pri-marily pair a single image with the corresponding textual report, limiting their ability to capture temporal relation-ships. To address this limitation, it is essential to learn temporal-aware cross-modal representations from sequen-tial medical images and text reports that highlight the tem-poral changes occurring between examinations. Specifi-cally, we introduce TempA- Vlp, a temporal-aware vision language pre-training framework with a cross-exam en-coder to integrate the information from both prior and cur-rent examinations. This approach enables the model to capture dynamic representations that reflect disease pro-gression over time, which allows us to (i) achieve state-of-the-art performance in disease progression classification, (ii) localize dynamic progression regions across consecutive examinations, as demonstrated in our new task, dynamic phrase grounding on the Chest-Imagenome Gold dataset, and (iii) highlight progression localized regions, often rele-vant to lesion areas, which in turn improves disease classi-fication tasks on a single image.

China’s digital collectibles utilize blockchain technology to create unique digital certificates for specific works and artworks that consumers can purchase, collect, and use. Digital collectibles are localized Chinese digital products, similar to non-fungible token (NFT), but with differences. This study explores the influencing factors and mechanisms of consumer purchase in digital collectibles. Our study developed a research model based on the Stimulus-Organism-Response (SOR) paradigm and the integration of perceived value theory and social identity theory. We used a questionnaire survey method and obtained 333 valid samples. Then, we used PLS-SEM to test the model. The empirical results are as follows: (1) Our study constructs and validates a multidimensional structure of perceived value and perceived risk; (2) consumers’ perceived value influences their purchase intentions through the mediating roles of consumer-digital collectibles cognitive identification and affective identification; (3) both supply-based scarcity and demand-based scarcity have a positive moderating effect on the relationship between cognitive identification and purchase intention. Our study reveals the pathway mechanism of the purchase intention and demonstrates the moderating roles of supply-based and demand-based scarcity. In addition, our study can provide practical suggestions for platforms and industries to conduct marketing activities and avoid risk.

Diffusion models achieved great success in image synthesis, but still face challenges in high-resolution generation. Through the lens of discrete cosine transformation, we find the main reason is that *the same noise level on a higher resolution results in a higher Signal-to-Noise Ratio in the frequency domain*. In this work, we present Relay Diffusion Model (RDM), which transfers a low-resolution image or noise into an equivalent high-resolution one for diffusion model via blurring diffusion and block noise. Therefore, the diffusion process can continue seamlessly in any new resolution or model without restarting from pure noise or low-resolution conditioning. RDM achieves state-of-the-art FID on CelebA-HQ and sFID on ImageNet 256$\times$256, surpassing previous works such as ADM, LDM and DiT by a large margin. All the codes and checkpoints are open-sourced at \url{https://github.com/THUDM/RelayDiffusion}.

We introduce CogVLM, a powerful open-source visual language foundation model. Different from the popular \emph{shallow alignment} method which maps image features into the input space of language model, CogVLM bridges the gap between the frozen pretrained language model and image encoder by a trainable visual expert module in the attention and FFN layers. As a result, CogVLM enables a deep fusion of vision language features without sacrificing any performance on NLP tasks. CogVLM-17B achieves state-of-the-art performance on 17 classic cross-modal benchmarks, including 1) image captioning datasets: NoCaps, Flicker30k, 2) VQA datasets: OKVQA, TextVQA, OCRVQA, ScienceQA, 3) LVLM benchmarks: MM-Vet, MMBench, SEED-Bench, LLaVABench, POPE, MMMU, MathVista, 4) visual grounding datasets: RefCOCO, RefCOCO+, RefCOCOg, Visual7W. Codes and checkpoints are available at Github.

In recent years, pretrained models revolutionized the paradigm of natural language understanding (NLU), where we append a randomly initialized classification head after the pretrained backbone, e.g.BERT, and finetune the whole model.As the pretrained backbone makes a major contribution to the improvement, we naturally expect a good pretrained classification head can also benefit the training.However, the final-layer output of the backbone, i.e. the input of the classification head, will change greatly during finetuning, making the usual head-only pretraining (LP-FT) ineffective.In this paper, we find that parameter-efficient tuning makes a good classification head, with which we can simply replace the randomly initialized heads for a stable performance gain.Our experiments demonstrate that the classification head jointly pretrained with parameter-efficient tuning consistently improves the performance on 9 tasks in GLUE and SuperGLUE.* Equal contribution.Codes are at https://github. com/THUDM/Efficient-Head-Finetuning.

Text-to-Image generation in the general domain has long been an open problem, which requires both a powerful generative model and cross-modal understanding. We propose CogView, a 4-billion-parameter Transformer with VQ-VAE tokenizer to advance this problem. We also demonstrate the finetuning strategies for various downstream tasks, e.g. style learning, super-resolution, text-image ranking and fashion design, and methods to stabilize pretraining, e.g. eliminating NaN losses. CogView achieves the state-of-the-art FID on the blurred MS COCO dataset, outperforming previous GAN-based models and a recent similar work DALL-E.

This paper studies distributed estimation and support recovery for high-dimensional linear regression model with heavy-tailed noise. To deal with heavy-tailed noise whose variance can be infinite, we adopt the quantile regression loss function instead of the commonly used squared loss. However, the non-smooth quantile loss poses new challenges to high-dimensional distributed estimation in both computation and theoretical development. To address the challenge, we transform the response variable and establish a new connection between quantile regression and ordinary linear regression. Then, we provide a distributed estimator that is both computationally and communicationally efficient, where only the gradient information is communicated at each iteration. Theoretically, we show that, after a constant number of iterations, the proposed estimator achieves a near-oracle convergence rate without any restriction on the number of machines. Moreover, we establish the theoretical guarantee for the support recovery. The simulation analysis is provided to demonstrate the effectiveness of our method.

The growing size of modern data brings many new challenges to existing statistical inference methodologies and theories, and calls for the development of distributed inferential approaches. This paper studies distributed inference for linear support vector machine (SVM) for the binary classification task. Despite a vast literature on SVM, much less is known about the inferential properties of SVM, especially in a distributed setting. In this paper, we propose a multi-round distributed linear-type (MDL) estimator for conducting inference for linear SVM. The proposed estimator is computationally efficient. In particular, it only requires an initial SVM estimator and then successively refines the estimator by solving simple weighted least squares problem. Theoretically, we establish the Bahadur representation of the estimator. Based on the representation, the asymptotic normality is further derived, which shows that the MDL estimator achieves the optimal statistical efficiency, i.e., the same efficiency as the classical linear SVM applying to the entire data set in a single machine setup. Moreover, our asymptotic result avoids the condition on the number of machines or data batches, which is commonly assumed in distributed estimation literature, and allows the case of diverging dimension. We provide simulation studies to demonstrate the performance of the proposed MDL estimator.