ViSMoE introduces a visual‑aware sparse Mixture‑of‑Experts framework for embodied referring expression grounding, enabling an agent to navigate real environments and localize a target object from natural language instructions. By routing visual information through specialized experts, the method produces discriminative representations for both navigation views and candidate objects, unlike prior approaches that use a single vision encoder. Experiments on the REVERIE and SOON datasets show that ViSMoE surpasses existing state‑of‑the‑art methods.
arXiv:2610.08401v1 Announce Type: cross Abstract: While recent vision-language models (VLMs) have shown outstanding performance across diverse applications, they tend to under-use visual information...
EPIC-Bench is a perception‑centric benchmark for fine‑grained embodied visual grounding in vision‑language models. It contains 6.6k human‑annotated tuples of Image, Text, and Mask across the three core stages of embodied interaction—Target Localization, Navigation, and Manipulation. The benchmark offers 23 fine‑grained tasks, multi‑dimensional settings, and evaluates over 80 leading VLMs to provide diagnostic insights into their vision‑language grounding capabilities.
arXiv:2607.10744v5 Announce Type: replace Abstract: Benefiting from the powerful priors embedded in large-scale pre-training data and the emerging commonsense reasoning ability, large language models...
arXiv:2607. 10383v1 Announce Type: cross Abstract: Visual Language Navigation foundation models aim to unify deep reasoning for grounded spatial decisions with broad versatility for diverse embodied tasks.
arXiv:2607. 09792v1 Announce Type: cross Abstract: Navigation is a fundamental capability of autonomous systems, yet most existing approaches rely on highly structured models and strong prior assumptions, limiting their robustness in open and uncertain real-world environments.
arXiv:2606. 15231v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) have demonstrated impressive capabilities in many visual tasks, but they often struggle with factual grounding when confronted with complex, open-world scenarios.
arXiv:2610.09440v1 Announce Type: new Abstract: Pre-trained vision encoders contain layer-wise visual representations that differ in spatial granularity, semantic abstraction, and sensitivity to loca...
The paper introduces a progressive training strategy for embodied vision‑language models aimed at reducing spatio‑temporal hallucinations. It first creates a Chain‑of‑Thought dataset that breaks complex reasoning into detailed spatiotemporal steps, then uses supervised pre‑training on this dataset followed by fine‑tuning with weakly‑labeled data. Experiments show the method improves backbone accuracy and narrows the forward‑backward performance gap from over 70% to 6.53%, indicating stronger dynamic reasoning and fewer temporal biases.
arXiv:2609.24576v1 Announce Type: cross Abstract: Modern Vision-Language Navigation (VLN) models rely mostly on pre-trained large Vision-Language Models (VLMs) to predict navigation actions. While th...
The paper introduces VGEBench, a new benchmark for evaluating Vision‑Language Models (VLMs) on generalizable, visually grounded exploration of household devices. Unlike existing datasets that rely on static images or annotated trajectories, VGEBench employs a logic‑driven state machine to simulate multi‑turn interaction loops, requiring agents to actively perceive, act, and refine their actions to achieve goals. Experiments show that current VLMs struggle to translate semantic knowledge into physical execution and to maintain long‑horizon state tracking.
arXiv:2604. 02327v2 Announce Type: replace-cross Abstract: Pretrained Vision Transformers (ViTs) such as DINOv2 and MAE provide generic image features that can be applied to a variety of downstream tasks such as retrieval, classification, and segmentation.
DobotTrack is a vision‑language‑action framework for embodied person tracking that uses auxiliary grounding tasks—person segmentation, bounding‑box grounding, visual‑spatial classification, future DINO feature prediction, and future‑frame prediction—to improve target delineation and localization. It introduces Q‑identity, a target‑conditioned historical visual compression module that preserves identity‑discriminative appearance features in compact tokens. On the EVT‑Bench dataset, the 0.8B‑parameter model achieves a 77.9% success rate, surpassing OmTrackVLA by 16.9 percentage points, and the authors also provide a real‑world quadruped‑robot dataset with 596,323 mask‑annotated frames for further training.