Facial analysis is a fundamental problem in vision–language research, with important applications in affective computing. However, existing methods primarily focus on global facial attributes or single-dimension analysis, lacking fine-grained, interpretable multi-attribute modeling of arbitrary local facial regions. We introduce FaceFocalDesc, a new problem that aims to generate and recognize multi-attribute natural language descriptions for arbitrarily selected facial regions. The target attributes include facial action units, emotional states, and age. We argue that explicit region-level modeling enables more controllable and interpretable facial understanding. To support this task, we construct a new dataset with region-level annotations and corresponding language descriptions. We further propose Focal-RegionFace, a vision–language model fine-tuned from Qwen2.5-VL, which progressively refines its focus on localized facial features through multi-stage training. Experiments show that Focal-RegionFace achieves state-of-the-art performance on the proposed benchmark under both standard and newly introduced metrics, demonstrating its effectiveness in fine-grained region-focused facial analysis.

Generative recommendation provides a novel paradigm in which each item is represented by a discrete semantic ID (SID) learned from rich content. Most methods treat SIDs as predefined and train recommenders under static indexing. In practice, SIDs are optimized only for content reconstruction rather than recommendation accuracy. This leads to an objective mismatch : the system optimizes an indexing loss to learn the SID, and a recommendation loss for interaction prediction, but because the tokenizer is trained independently, the recommendation loss cannot update it. A natural approach is to make semantic indexing differentiable so recommendation gradients can directly influence SID learning, but this often causes codebook collapse with only a few codes used. We attribute this to early deterministic assignments that limit codebook exploration, leading to imbalance and unstable optimization. In this paper, we therefore propose DIGER (Differentiable Semantic ID for GEnerative Recommendation). DIGER is a first step towards an effective differentiable semantic ID for generative recommendation. The Gumbel noise explicitly encourages early-stage exploration over codes, mitigating collapse and improving code utilization. To better balance exploration and convergence, we introduce two uncertainty decay strategies that reduce the Gumbel noise, enabling a gradual shift from early-stage exploration to the exploitation of learned SIDs. Extensive experiments across multiple public datasets demonstrate consistent improvements from differentiable semantic ID. These results confirm the effectiveness of aligning indexing and recommendation objectives through differentiable SIDs. This identifies differentiable SID as a promising area of study. Our code is released under https://github.com/junchen-fu/DIGER.

Sequential Recommender Systems (SRSs) predict items of interest for users based on their historical interactions. Two key popular paradigms for SRs are unidirectional Next Item Prediction (NIP) and bidirectional Masked Language Modeling (MLM). NIP performs well in recommendation tasks but is constrained by its reliance on prior information and a rigid temporal order assumption, limiting its ability to capture dynamic personalized preferences. MLM, on the other hand, enhances sequence representation and personalization by leveraging bidirectional context, but its objective is misaligned with recommendation tasks. To this end, we develop MetaSR, a meta-learning-based SR approach that ingeniously combines both paradigms into a unified framework to enhance personalized sequential recommendation. Specifically, we treat the mask prediction task in MLM as the inner-loop task (i.e., the support set), adjusting mask items dynamically to generate sequences that match individual behavioral patterns to capture personalized user preferences. In the outer loop (i.e., the query set), we adopt the NIP paradigm to directly train MetaSR toward the SR target. The joint training bridges the gap between MLM and SRs tasks, also overcoming NIP’s limitations in capturing dynamic personalized preferences. Furthermore, to improve the overall efficiency, we develop a reinforcement learning-based Adaptive Masked Sequence Selection (AMSS) mechanism, automatically selecting the optimal masking prediction tasks within the meta-learning support set to accelerate the meta-learning and achieve better personalization. Experiments on multiple public benchmark datasets show that MetaSR significantly outperforms existing SRs models. The dataset and codes are available at MetaSR .

Multimodal recommender systems (MMRec) leverage multimodal features, such as visual and textual data, to improve recommendation performance, playing a key role in platforms like online shopping and short videos. However, the large modality encoders and complex processing modules of MMRec significantly reduce its efficiency. A promising solution is compressing MMRec into an ID-based MLP model (MLPRec), which has a simpler structure and avoids complex modality handling. However, traditional knowledge distillation methods struggle to transfer knowledge effectively from MMRec to MLPRec, due to differences in their model structure and capacity. To address this, we propose a frequency-decoupled knowledge distillation framework-FDRec-to efficiently transfer knowledge from MMRec to MLPRec. By analyzing graph signals from a signal processing perspective, we propose decoupling the distillation process into low-frequency and high-frequency components, ensuring effective transmission of challenging high-frequency knowledge while preventing it from being overshadowed by monotonous low-frequency signals. To address the instability and fragmentation issues of KL divergence in traditional distillation approaches, we introduce the Wasserstein distance, which captures geometric structure and provides stable gradients. Additionally, FDRec incorporates an embedding-level contrastive learning method, further enhancing the transfer of refined knowledge from MMRec and injecting graph structure information into MLPRec for more effective distillation. Extensive experiments on four benchmark datasets and five popular MMRec models show that FDRec not only significantly reduces the computational costs and improves the inference efficiency, but also achieves comparable or even superior performance compared to MMRec. Our code is available at: https://github.com/Suehn/FDRec_

Video generation assessment is essential for ensuring that generative models produce visually realistic, high-quality videos while aligning with human expectations. Current video generation benchmarks fall into two main categories: traditional benchmarks, which use metrics and embeddings to evaluate generated video quality across multiple dimensions but often lack alignment with human judgments; and large language model (LLM)-based benchmarks, though capable of human-like reasoning, are constrained by a limited understanding of video quality metrics and cross-modal consistency. To address these challenges and establish a benchmark that better aligns with human preferences, this paper introduces Video-Bench, a comprehensive benchmark featuring a rich prompt suite and extensive evaluation dimensions. This benchmark represents the first attempt to systematically leverage MLLMs across all dimensions relevant to video generation assessment in generative models. By incorporating few-shot scoring and chain-of-query techniques, Video-Bench provides a structured, scalable approach to generated video evaluation. Experiments on advanced models including Sora demonstrate that Video-bench achieve superior alignment with human preferences across all dimensions. Moreover, in instances where our framework’s assessments diverge from human evaluations, it consistently offers more objective and accurate insights, suggesting an even greater potential advantage over traditional human judgment.

Multimodal foundation models have significantly improved feature representation by integrating information from multiple modalities, making them highly suitable for a broader set of applications. However, the exploration of multimodal facial representation for understanding perception has been limited. Understanding and analyzing facial states, such as Action Units (AUs) and emotions, require a comprehensive and robust framework that bridges visual and linguistic modalities. In this paper, we present a comprehensive pipeline for multimodal facial state analysis. First, we compile a new Multimodal Face Dataset (MFA) by generating detailed multilevel language descriptions of face, incorporating Action Unit (AU) and emotion descriptions, by leveraging GPT-4o. Second, we introduce a novel Multilevel Multimodal Face Foundation model (MF2) tailored for Action Unit (AU) and emotion recognition. Our model incorporates comprehensive visual feature modeling at both local and global levels of face image, enhancing its ability to represent detailed facial appearances. This design aligns visual representations with structured AU and emotion descriptions, ensuring effective cross-modal integration. Third, we develop a Decoupled Fine-Tuning Network (DFN) that efficiently adapts MF2across various tasks and datasets. This approach not only reduces computational overhead but also broadens the applicability of the foundation model to diverse scenarios. Experimentation show superior performance for AU and emotion detection tasks.

In this paper, we present a novel approach, termed Double-Filter,to “slim down” the fine-tuning process of vision-language pre-trained (VLP) models via filtering redundancies in feature inputs and architectural components. We enhance the fine-tuning process using two approaches. First, we develop a new patch selection method incorporating image patch filtering through background and foreground separation, followed by a refined patch selection process. Second, we design a genetic algorithm to eliminate redundant fine-grained architecture layers, improving the efficiency and effectiveness of the model. The former makes patch selection semantics more comprehensive, improving inference efficiency while ensuring semantic representation. The latter’s fine-grained layer filter removes architectural redundancy to the extent possible and mitigates the impact on performance. Experimental results demonstrate that the proposed Double-Filter achieves superior efficiency of model fine-tuning and maintains competitive performance compared with the advanced efficient fine-tuning methods on three downstream tasks, VQA, NLVR and Retrieval. In addition, it has been proven to be effective under METER and ViLT VLP models.

The surge in multimedia content has led to the development of Multi-Modal Recommender Systems (MMRecs), which use diverse modalities-like text, images, videos, and audio-for more personalized recommendations. However, MMRecs struggle with the challenge of noisy data from the misalignment among modal content and the difference between modal and recommendation semantics, while traditional denoising methods fall short in addressing these issues due to the complexity of multi-modal data. To overcome this, we propose a universal guided in-sync distillation denoising framework for multi-modal recommndation (GUIDER), aimed at improving MMRecs by denoising user feedbacks. Specifically, GUIDER employs a re-calibration strategy to identify clean and noisy interactions from modal content. Furthermore, it incorporates a Denoising Bayesian Personalized Ranking (DBPR) loss function to denoise implicit user feedback. Finally, it utilizes a denoising knowledge distillation objective based on Optimal Transport (OT) distance to guide the mapping from modality representations to recommendation semantics spaces. GUIDER can be seamlessly integrated into existing MMRecs methods as a plug-and-play solution for recommendation denoising. Experiment results on four public datasets show its effectiveness and universality across various MMRecs.

Multi-modal recommender systems (MMRecs) leverage diverse modalities to deliver personalized recommendations, yet they often struggle with efficiency due to the large size of modality encoders and the complexity of fusing high-dimensional features. To address the efficiency issue, a promising solution is to compress a cumbersome MMRec into a lightweight ID-based Multi- Layer Perceptron-based Recommender system (MLPRec) through Knowledge Distillation (KD). Despite effectiveness, we argue that this approach overlooks the significant gap between the complex teacher MMRec and the lightweight, ID-based student MLPRec, which differ significantly in size, architecture, and input modalities, leading to ineffective knowledge transfer and suboptimal student performance. To bridge this gap, we propose TARec, a novel teacher-assisted Wasserstein Knowledge Distillation framework for compressing MMRecs into an efficient MLPRec. TARec introduces: (i) a two-staged KD process using an intermediate Teacher Assistant (TA) model to bridge the gap between teacher and student, facilitating smoother knowledge transfer; (ii) logit-level KD using the Wasserstein Distance as metric, replacing the conventional KL divergence to ensure stable gradient flow even with significant teacher-student gaps; and (iii) embedding-level contrastive KD to further distill high-quality embedding-level knowledge from teacher. Extensive experiments on real-world datasets verify the effectiveness of TARec, demonstrating that TARec significantly outperforms the state-of-the-art MMRecs while reducing computational costs. Our code is available at: https://github.com/Suehn/TARec.git.

Large foundational models, through upstream pre-training and downstream fine-tuning, have achieved immense success in the broad AI community due to improved model performance and significant reductions in repetitive engineering. By contrast, the transferable one-for-all models in the recommender system field, referred to as TransRec, have made limited progress. The development of TransRec has encountered multiple challenges, among which the lack of large-scale, high-quality transfer learning recommendation dataset and benchmark suites is one of the biggest obstacles. To this end, we introduce NineRec, a TransRec dataset suite that comprises a large-scale source domain recommendation dataset and nine diverse target domain recommendation datasets. Each item in NineRec is accompanied by a descriptive text and a high-resolution cover image. Leveraging NineRec, we enable the implementation of TransRec models by learning from raw multimodal features instead of relying solely on pre-extracted off-the-shelf features. Finally, we present robust TransRec benchmark results with several classical network architectures, providing valuable insights into the field.

Multimodal foundation models (MFMs) have revolutionized sequential recommender systems through advanced representation learning. While Parameter-efficient Fine-tuning (PEFT) is commonly used to adapt these models, studies often prioritize parameter efficiency, neglecting GPU memory and training speed. To address this, we introduced the IISAN framework, significantly enhancing efficiency. However, IISAN was limited to symmetrical MFMs and identical text and image encoders, preventing the use of state-of-the-art Large Language Models. To overcome this, we developed IISAN-Versa, a versatile plug-and-play architecture compatible with both symmetrical and asymmetrical MFMs. IISAN-Versa employs a Decoupled PEFT structure and utilizes both intra- and inter-modal adaptation. It effectively handles asymmetry through a simple yet effective combination of group layer-dropping and dimension transformation alignment. Our research demonstrates that IISAN-Versa effectively adapts large text encoders, and we further identify a scaling effect where larger text encoders generally perform better. IISAN-Versa also demonstrates strong versatility in our defined multimodal scenarios, which include raw titles and captions generated from images and videos. Additionally, IISAN-Versa achieved state-of-the-art performance on the MicroLens public benchmark.

Facial action units (AUs), as defined in the Facial Action Coding System (FACS), have received significant research interest owing to their diverse range of applications in facial state analysis. Current mainstream FAU recognition models have a notable limitation, i.e., focusing only on the accuracy of AU recognition and overlooking explanations of corresponding AU states. In this paper, we propose an end-to-end Vision-Language joint learning network for explainable FAU recognition (termed VL-FAU), which aims to reinforce AU representation capability and language interpretability through the integration of joint multimodal tasks. Specifically, VL-FAU brings together language models to generate fine-grained local muscle descriptions and distinguishable global face description when optimising FAU recognition. Through this, the global facial representation and its local AU representations will achieve higher distinguishability among different AUs and different subjects. In addition, multi-level AU representation learning is utilised to improve AU individual attention-aware representation capabilities based on multi-scale combined facial stem feature. Extensive experiments on DISFA and BP4D AU datasets show that the proposed approach achieves superior performance over the state-of-the-art methods on most of the metrics. In addition, compared with mainstream FAU recognition methods, VL-FAU can provide local- and global-level interpretability language descriptions with the AUs' predictions.

Multimodal foundation models are transformative in sequential recommender systems, leveraging powerful representation learning capabilities. While Parameter-efficient Fine-tuning (PEFT) is commonly used to adapt foundation models for recommendation tasks, most research prioritizes parameter efficiency, often overlooking critical factors like GPU memory efficiency and training speed. Addressing this gap, our paper introduces IISAN (Intra- and Inter-modal Side Adapted Network for Multimodal Representation), a simple plug-and-play architecture using a Decoupled PEFT structure and exploiting both intra- and inter-modal adaptation. IISAN matches the performance of full fine-tuning (FFT) and state-of-the-art PEFT. More importantly, it significantly reduces GPU memory usage - from 47GB to just 3GB for multimodal sequential recommendation tasks. Additionally, it accelerates training time per epoch from 443s to 22s compared to FFT. This is also a notable improvement over the Adapter and LoRA, which require 37-39 GB GPU memory and 350-380 seconds per epoch for training. Furthermore, we propose a new composite efficiency metric, TPME (Training-time, Parameter, and GPU Memory Efficiency) to alleviate the prevalent misconception that "parameter efficiency represents overall efficiency". TPME provides more comprehensive insights into practical efficiency comparisons between different methods. Besides, we give an accessible efficiency analysis of all PEFT and FFT approaches, which demonstrate the superiority of IISAN. We release our codes and other materials at https://github.com/GAIR-Lab/IISAN.

Adapters, a plug-in neural network module with some tunable parameters, have emerged as a parameter-efficient transfer learning technique for adapting pre-trained models to downstream tasks, especially for natural language processing (NLP) and computer vision (CV) fields. Meanwhile, learning recommendation models directly from raw item modality features --- e.g., texts of NLP and images of CV --- can enable effective and transferable recommender systems (called TransRec). In view of this, a natural question arises:can adapter-based learning techniques achieve parameter-efficient TransRec with good performance?

Recommendation models that utilize unique identities (IDs for short) to represent distinct users and items have been state-of-the-art (SOTA) and dominated the recommender systems (RS) literature for over a decade. Meanwhile, the pre-trained modality encoders, such as BERT [9] and Vision Transformer [11], have become increasingly powerful in modeling the raw modality features of an item, such as text and images. Given this, a natural question arises: can a purely modality-based recommendation model (MoRec) outperforms or matches a pure ID-based model (IDRec) by replacing the itemID embedding with a SOTA modality encoder? In fact, this question was answered ten years ago when IDRec beats MoRec by a strong margin in both recommendation accuracy and efficiency.