Yuhang Li

Kangsheng Wang, Yuhang Li, Chengwei Ye et al.

Apparent personality analysis from short videos poses significant chal-lenges due to the complex interplay of visual, auditory, and textual cues. In this paper, we propose GAME, a Graph-Augmented Multimodal Encoder designed to robustly model and fuse multi-source features for automatic personality prediction. For the visual stream, we construct a facial graph and introduce a dual-branch Geo Two-Stream Network, which combines Graph Convolutional Networks (GCNs) and Convolutional Neural Net-works (CNNs) with attention mechanisms to capture both structural and appearance-based facial cues. Complementing this, global context and iden-tity features are extracted using pretrained ResNet18 and VGGFace back-bones. To capture temporal dynamics, frame-level features are processed by a BiGRU enhanced with temporal attention modules. Meanwhile, audio representations are derived from the VGGish network, and linguistic se-mantics are captured via the XLM-Roberta transformer. To achieve effective multimodal integration, we propose a Channel Attention-based Fusion module, followed by a Multi-Layer Perceptron (MLP) regression head for predicting personality traits. Extensive experiments show that GAME con-sistently outperforms existing methods across multiple benchmarks, vali-dating its effectiveness and generalizability.

Yangfu Li, Hongjian Zhan, Tianyi Chen et al.

Existing visual token pruning methods target prompt alignment and visual preservation with static strategies, overlooking the varying relative importance of these objectives across tasks, which leads to inconsistent performance. To address this, we derive the first closed-form error bound for visual token pruning based on the Hausdorff distance, uniformly characterizing the contributions of both objectives. Moreover, leveraging $ε$-covering theory, we reveal an intrinsic trade-off between these objectives and quantify their optimal attainment levels under a fixed budget. To practically handle this trade-off, we propose Multi-Objective Balanced Covering (MoB), which reformulates visual token pruning as a bi-objective covering problem. In this framework, the attainment trade-off reduces to budget allocation via greedy radius trading. MoB offers a provable performance bound and linear scalability with respect to the number of input visual tokens, enabling adaptation to challenging pruning scenarios. Extensive experiments show that MoB preserves 96.4% of performance for LLaVA-1.5-7B using only 11.1% of the original visual tokens and accelerates LLaVA-Next-7B by 1.3-1.5$\times$ with negligible performance loss. Additionally, evaluations on Qwen2-VL and Video-LLaVA confirm that MoB integrates seamlessly into advanced MLLMs and diverse vision-language tasks.