Handshape and orientation are geometric components of sign language, yet many sign encoders learn from appearance alone. We introduce SignDEG (DINO Eats Geometry), a temporal sign encoder that uses articulated hand geometry as a pre-training target and requires only video at inference. Frozen DINOv3 features of tracked...
Jun-Yi Hu, Zhe-Wen He, Hao Huang et al.· 0 citations
Visual interventions provide a training-free way to reduce hallucinations in multimodal large language models. However, an intervention may have little leverage at some layers and may damage a correct answer even when it changes the output. We investigate where visual interventions retain leverage and when they improve...
Qi-Ji Zhou, Jun-Yi Hu, Pei-Yu Wang et al.· 0 citations
A robot demonstration records two things in the same frames: what happened to the objects, and how one particular arm made it happen. We condition on the first. A demonstration is compiled into an effect program: the 3D keypoint trajectories of the objects that moved, two points marking where each was held, and the con...
Jun-Yi Hu, Zhe-Wen He, Zhen-Hua Li et al.· 0 citations
Power sampling has emerged as a training-free approach to LLM reasoning, eliciting capabilities comparable to reinforcement learning by sharpening the model distribution over complete responses. Despite this success, power sampling remains underexplored in large vision-language models (LVLMs). We transfer Power-SMC to...
Yao Zhang, Xiang-Yu Qiu, Jun-Yi Hu et al.· 0 citations
Referring Expression Comprehension (REC) is commonly studied under dataset-specific fine-tuning, resulting in specialist models with limited cross-dataset generalization. In this work, we revisit REC from the perspective of unified open-vocabulary grounding and identify representation degeneration as a key obstacle to...
Jun-Yi Hu, Tian Bai, Feng-Yi Wu et al.· 0 citations
This work introduces LAGO (LAnguage-Guided adaptive Object-region focus), which reframes localized recognition as language-guided directed region discovery and addresses the circular dependence between recognizing a class and locating its supporting evidence, while preserving complementary local, contextual, and global...
Jun-Yi Hu, Qiji Zhou, Lei Zhang et al.· arXiv.org· 0 citations
Embodied visual tracking requires a robot to choose actions that keep a moving target observable at a suitable distance, and to recover it after occlusion, out-of-view drift, or distractor crossings. We cast the task as planning over future target evidence with an action-conditioned world model. In logged tracking data...
Jun-Yi Hu, Shuaihang Yuan, Jia-Zhao Liang et al.· 0 citations
SignDino, a self-supervised sign-video encoder that moves the DINOv3 student--teacher recipe from the spatial domain of image crops to the temporal domain of tracked sign streams, provides a strong public self-supervised representation and shows competitive or state-of-the-art performance under matched downstream evalu...
Jun-Yi Hu, Zhe-Wen He, Hao Huang et al.· 0 citations
VTaMo is presented, a framework that introduces explicit multi-granularity alignment at three levels: local alignment via entropy-regularized optimal transport with a learnable null token for fine-grained frame-to-token correspondences; global alignment via a learnable orthogonal transformation that calibrates embeddin...
