Extending Visual Dynamics for Video-to-Music Generation

Music profoundly enhances video production by improving quality, engagement, and emotional resonance, sparking growing interest in video-to-music generation. Despite recent advances, existing approaches remain limited in specific scenarios or undervalue the visual dynamics. To address these limitations, we focus on tackling the complexity of dynamics and resolving temporal misalignment between video and music representations. To this end, we propose DyViM, a novel framework to enhance dynamics modeling for video-to-music generation. Specifically, we extract frame-wise dynamics features via a simplified motion encoder inherited from optical flow methods, followed by a self-attention module for aggregation within frames. These dynamic features are then incorporated to extend existing music tokens for temporal alignment. Additionally, high-level semantics are conveyed through a cross-attention mechanism, and an annealing tuning strategy benefits to fine-tune well-trained music decoders efficiently, therefore facilitating seamless adaptation. Extensive experiments demonstrate DyViM's superiority over state-of-the-art (SOTA) methods.

Benchmarking Multimodal CoT Reward Model Stepwise by Visual Program

Recent advancements in reward signal usage for Large Language Models (LLMs) are remarkable. However, significant challenges exist when transitioning reward signal to the multimodal domain, including labor-intensive annotations, over-reliance on one-step rewards, and inadequate evaluation. To address these issues, we propose SVIP, a novel approach to train a step-level multi-dimensional Chain-of-Thought~(CoT) reward model automatically. It generates code for solving visual tasks and transforms the analysis of code blocks into the evaluation of CoT step as training samples. Then, we train SVIP-Reward model using a multi-head attention mechanism called TriAtt-CoT. The advantages of SVIP-Reward are evident throughout the entire process of MLLM. We also introduce a benchmark for CoT reward model training and testing. Experimental results demonstrate that SVIP-Reward improves MLLM performance across training and inference-time scaling, yielding better results on benchmarks while reducing hallucinations and enhancing reasoning ability.

LightPROF: A Lightweight Reasoning Framework for Large Language Model on Knowledge Graph

Large Language Models (LLMs) have impressive capabilities in text understanding and zero-shot reasoning. However, delays in knowledge updates may cause them to reason incorrectly or produce harmful results. Knowledge Graphs (KGs) provide rich and reliable contextual information for the reasoning process of LLMs by structurally organizing and connecting a wide range of entities and relations. Existing KG-based LLM reasoning methods only inject KGs' knowledge into prompts in a textual form, ignoring its structural information. Moreover, they mostly rely on close-source models or open-source models with large parameters, which poses challenges to high resource consumption. To address this, we propose a novel Lightweight and efficient Prompt learning-ReasOning Framework for KGQA (LightPROF), which leverages the full potential of LLMs to tackle complex reasoning tasks in a parameter-efficient manner. Specifically, LightPROF follows a "Retrieve-Embed-Reason process", first accurately, and stably retrieving the corresponding reasoning graph from the KG through retrieval module. Next, through a Transformer-based Knowledge Adapter, it finely extracts and integrates factual and structural information from the KG, then maps this information to the LLM's token embedding space, creating an LLM-friendly prompt to be used by the LLM for the final reasoning. Additionally, LightPROF only requires training Knowledge Adapter and can be compatible with any open-source LLM. Extensive experiments on two public KGQA benchmarks demonstrate that LightPROF achieves superior performance with small-scale LLMs. Furthermore, LightPROF shows significant advantages in terms of input token count and reasoning time.

Any2Caption:Interpreting Any Condition to Caption for Controllable Video Generation

To address the bottleneck of accurate user intent interpretation within the current video generation community, we present Any2Caption, a novel framework for controllable video generation under any condition. The key idea is to decouple various condition interpretation steps from the video synthesis step. By leveraging modern multimodal large language models (MLLMs), Any2Caption interprets diverse inputs--text, images, videos, and specialized cues such as region, motion, and camera poses--into dense, structured captions that offer backbone video generators with better guidance. We also introduce Any2CapIns, a large-scale dataset with 337K instances and 407K conditions for any-condition-to-caption instruction tuning. Comprehensive evaluations demonstrate significant improvements of our system in controllability and video quality across various aspects of existing video generation models. Project Page: https://sqwu.top/Any2Cap/

JavisDiT: Joint Audio-Video Diffusion Transformer with Hierarchical Spatio-Temporal Prior Synchronization

This paper introduces JavisDiT, a novel Joint Audio-Video Diffusion Transformer designed for synchronized audio-video generation (JAVG). Built upon the powerful Diffusion Transformer (DiT) architecture, JavisDiT is able to generate high-quality audio and video content simultaneously from open-ended user prompts. To ensure optimal synchronization, we introduce a fine-grained spatio-temporal alignment mechanism through a Hierarchical Spatial-Temporal Synchronized Prior (HiST-Sypo) Estimator. This module extracts both global and fine-grained spatio-temporal priors, guiding the synchronization between the visual and auditory components. Furthermore, we propose a new benchmark, JavisBench, consisting of 10,140 high-quality text-captioned sounding videos spanning diverse scenes and complex real-world scenarios. Further, we specifically devise a robust metric for evaluating the synchronization between generated audio-video pairs in real-world complex content. Experimental results demonstrate that JavisDiT significantly outperforms existing methods by ensuring both high-quality generation and precise synchronization, setting a new standard for JAVG tasks. Our code, model, and dataset will be made publicly available at https://javisdit.github.io/.

Exploring Training and Inference Scaling Laws in Generative Retrieval

Generative retrieval has emerged as a novel paradigm that leverages large language models (LLMs) to autoregressively generate document identifiers. Although promising, the mechanisms that underpin its performance and scalability remain largely unclear. We conduct a systematic investigation of training and inference scaling laws in generative retrieval, exploring how model size, training data scale, and inference-time compute jointly influence retrieval performance. To address the lack of suitable metrics, we propose a novel evaluation measure inspired by contrastive entropy and generation loss, providing a continuous performance signal that enables robust comparisons across diverse generative retrieval methods. Our experiments show that n-gram-based methods demonstrate strong alignment with both training and inference scaling laws, especially when paired with larger LLMs. Furthermore, increasing inference computation yields substantial performance gains, revealing that generative retrieval can significantly benefit from higher compute budgets at inference. Across these settings, LLaMA models consistently outperform T5 models, suggesting a particular advantage for larger decoder-only models in generative retrieval. Taken together, our findings underscore that model sizes, data availability, and inference computation interact to unlock the full potential of generative retrieval, offering new insights for designing and optimizing future systems.

Boosting Virtual Agent Learning and Reasoning: A Step-wise, Multi-dimensional, and Generalist Reward Model with Benchmark

The development of Generalist Virtual Agents (GVAs) powered by Multimodal Large Language Models (MLLMs) has shown significant promise in autonomous task execution. However, current training paradigms face critical limitations, including reliance on outcome supervision and labor-intensive human annotations. To address these challenges, we propose Similar, a Step-wise Multi-dimensional Generalist Reward Model, which offers fine-grained signals for agent training and can choose better action for inference-time scaling. Specifically, we begin by systematically defining five dimensions for evaluating agent actions. Building on this framework, we design an MCTS-P algorithm to automatically collect and annotate step-wise, five-dimensional agent execution data. Using this data, we train Similar with the Triple-M strategy. Furthermore, we introduce the first benchmark in the virtual agent domain for step-wise, multi-dimensional reward model training and evaluation, named SRM. This benchmark consists of two components: SRMTrain, which serves as the training set for Similar, and SRMEval, a manually selected test set for evaluating the reward model. Experimental results demonstrate that Similar, through its step-wise, multi-dimensional assessment and synergistic gain, provides GVAs with effective intermediate signals during both training and inference-time scaling. The code is available at https://github.com/Galery23/Similar-v1.

Towards Unified Latent Space for 3D Molecular Latent Diffusion Modeling

3D molecule generation is crucial for drug discovery and material science, requiring models to process complex multi-modalities, including atom types, chemical bonds, and 3D coordinates. A key challenge is integrating these modalities of different shapes while maintaining SE(3) equivariance for 3D coordinates. To achieve this, existing approaches typically maintain separate latent spaces for invariant and equivariant modalities, reducing efficiency in both training and sampling. In this work, we propose \textbf{U}nified Variational \textbf{A}uto-\textbf{E}ncoder for \textbf{3D} Molecular Latent Diffusion Modeling (\textbf{UAE-3D}), a multi-modal VAE that compresses 3D molecules into latent sequences from a unified latent space, while maintaining near-zero reconstruction error. This unified latent space eliminates the complexities of handling multi-modality and equivariance when performing latent diffusion modeling. We demonstrate this by employing the Diffusion Transformer--a general-purpose diffusion model without any molecular inductive bias--for latent generation. Extensive experiments on GEOM-Drugs and QM9 datasets demonstrate that our method significantly establishes new benchmarks in both \textit{de novo} and conditional 3D molecule generation, achieving leading efficiency and quality.

Continual Multimodal Contrastive Learning

Multimodal contrastive learning (MCL) advances in aligning different modalities and generating multimodal representations in a joint space. By leveraging contrastive learning across diverse modalities, large-scale multimodal data enhances representational quality. However, a critical yet often overlooked challenge remains: multimodal data is rarely collected in a single process, and training from scratch is computationally expensive. Instead, emergent multimodal data can be used to optimize existing models gradually, \textit{i.e.}, models are trained on a sequence of modality pair data. We define this problem as Continual Multimodal Contrastive Learning (CMCL), an underexplored yet crucial research direction at the intersection of multimodal and continual learning. In this paper, we formulate CMCL through two specialized principles of stability and plasticity. We theoretically derive a novel optimization-based method, which projects updated gradients from dual sides onto subspaces where any gradient is prevented from interfering with the previously learned knowledge. Two upper bounds provide theoretical insights on both stability and plasticity in our solution. Beyond our theoretical contributions, we conduct experiments on multiple datasets by comparing our method against advanced continual learning baselines. The empirical results further support our claims and demonstrate the efficacy of our method. The code will be publicly available.

Universal Scene Graph Generation

Scene graph (SG) representations can neatly and efficiently describe scene semantics, which has driven sustained intensive research in SG generation. In the real world, multiple modalities often coexist, with different types, such as images, text, video, and 3D data, expressing distinct characteristics. Unfortunately, current SG research is largely confined to single-modality scene modeling, preventing the full utilization of the complementary strengths of different modality SG representations in depicting holistic scene semantics. To this end, we introduce Universal SG (USG), a novel representation capable of fully characterizing comprehensive semantic scenes from any given combination of modality inputs, encompassing modality-invariant and modality-specific scenes. Further, we tailor a niche-targeting USG parser, USG-Par, which effectively addresses two key bottlenecks of cross-modal object alignment and out-of-domain challenges. We design the USG-Par with modular architecture for end-to-end USG generation, in which we devise an object associator to relieve the modality gap for cross-modal object alignment. Further, we propose a text-centric scene contrasting learning mechanism to mitigate domain imbalances by aligning multimodal objects and relations with textual SGs. Through extensive experiments, we demonstrate that USG offers a stronger capability for expressing scene semantics than standalone SGs, and also that our USG-Par achieves higher efficacy and performance.