Searcharxiv⌕ Search

arXiv subjects

Lingyun Li

Publications and source records attributed to Lingyun Li.

8 recordsLinked to original sources
arXiv2026

SpaceMind++: Toward Allocentric Cognitive Maps for Spatially Grounded Video MLLMs

Recent multimodal large language models (MLLMs) have made remarkable progress in visual understanding and language-based reasoning, yet they lack a persistent world-centered representation for spatially consistent reasoning in 3D environments. Inspired by the mammalian dual-stream system, where semantic and spatial cues are processed separately and integrated into an allocentric cognitive map, we propose SpaceMind++, a video MLLM architecture that explicitly builds a voxelized cognitive map from RGB videos. This map reorganizes fragmented egocentric observations into a shared 3D metric representation, enabling the model to preserve object permanence and spatial topology across changing viewpoints. To make this allocentric representation usable by a pretrained video MLLM without disrupting its native visual-token interface, we introduce Coordinate-Guided Deep Iterative Fusion, a new mechanism that relays map-level spatial knowledge back into the original 2D visual features. This fusion is explicitly guided by coordinate embeddings and 3D Rotary Positional Encoding, which ground semantic interactions in metric 3D space, resembling the entorhinal binding of sensory features to metric space. Extensive experiments show that SpaceMind++ achieves new state-of-the-art performance on VSI-Bench. Furthermore, it demonstrates superior out-of-distribution generalization on SPBench, SITE-Bench, and SPAR-Bench, underscoring its robustness in unseen 3D environments.

cs.CV↗
arXiv2026

Token Entropy Regularization for Multi-modal Antenna Affiliation Identification

Accurate antenna affiliation identification is crucial for optimizing and maintaining communication networks. Current practice, however, relies on the cumbersome and error-prone process of manual tower inspections. We propose a novel paradigm shift that fuses video footage of base stations, antenna geometric features, and Physical Cell Identity (PCI) signals, transforming antenna affiliation identification into multi-modal classification and matching tasks. Publicly available pretrained transformers struggle with this unique task due to a lack of analogous data in the communications domain, which hampers cross-modal alignment. To address this, we introduce a dedicated training framework that aligns antenna images with corresponding PCI signals. To tackle the representation alignment challenge, we propose a novel Token Entropy Regularization module in the pretraining stage. Our experiments demonstrate that TER accelerates convergence and yields significant performance gains. Further analysis reveals that the entropy of the first token is modality-dependent. Code will be made available upon publication.

cs.CV↗
arXiv2025

SpaceMind: Camera-Guided Modality Fusion for Spatial Reasoning in Vision-Language Models

Large vision-language models (VLMs) show strong multimodal understanding but still struggle with 3D spatial reasoning, such as distance estimation, size comparison, and cross-view consistency. Existing 3D-aware methods either depend on auxiliary 3D information or enhance RGB-only VLMs with geometry encoders through shallow feature fusion. We propose SpaceMind, a multimodal large language model explicitly designed for spatial reasoning solely from RGB inputs. The model adopts a dual-encoder architecture, integrating VGGT as a spatial understanding encoder and InternViT as a 2D visual encoder. The key idea is to treat the camera representation as an active guiding modality rather than passive metadata. Specifically, SpaceMind introduces a lightweight Camera-Guided Modality Fusion module before the language model to replace shallow fusion. It applies camera-conditioned biasing to spatial tokens, assigns query-independent weights reflecting their geometric importance, and uses the camera embedding to gate the fused representation. Empirically, SpaceMind establishes new state-of-the-art results on VSI-Bench, SQA3D and SPBench, surpassing both open and proprietary systems on VSI-Bench and SPBench by large margins and achieving state-of-the-art performance on SQA3D. These results demonstrate that camera-guided modality fusion is an effective and practical inductive bias for equipping VLMs with genuinely spatially grounded intelligence. We will release code and model checkpoints to support future research.

cs.CV↗
arXiv2023

Molecular tuning of DNA framework-programmed silicification by cationic silica cluster attachment

The organizational complexity of biominerals has long fascinated scientists seeking to understand biological programming and implement new developments in biomimetic materials chemistry. Nonclassical crystallization pathways have been observed and analyzed in typical crystalline biominerals, involving the controlled attachment and reconfiguration of nanoparticles and clusters on organic templates. However, the understanding of templated amorphous silica mineralization remains limited, hindering the rational design of complex silica-based materials. Here, we present a systematic study on the stabilization of self-capping cationic silica cluster (CSC) and their assembly dynamics using DNA nanostructures as programmable attachment templates. By tuning the composition and structure of CSC, we demonstrate high-fidelity silicification at single-cluster resolution, revealing a process of adaptive templating involving cooperative adjustments of both the DNA framework and cluster morphology. Our results provide a unified model of silicification by cluster attachment and pave the way towards the molecular tuning of pre- and post-nucleation stages of sol-gel reactions. Overall, our findings provide new insights for the design of silica-based materials with controlled organization and functionality, bridging the gap between biomineralization principles and the rational design of biomimetic material.

physics.chem-ph↗
arXiv2023

High dynamic range open-loop current measurement based on diamond quantum magnetometer achieving ppm scale precision

Negatively charged nitrogen-vacancy (NV) centers in diamond have been extensively studied as a promising high sensitivity solid-state magnetic field sensor at room temperature. However, their use for current sensing applications is limited due to the challenge of integration and miniaturization of the diamond NV sensor. Here, we demonstrate an integrated NV sensor fabricated with standard microfabrication process. The sensor device incorporated with a toroidal magnetic yoke enables a high-precision wide range direct current sensing with galvanic isolation. The performance of the diamond NV current sensor in an open loop configuration has been investigated. A current measuring range of 0 A ~ 1000 A with an uncertainty of 46 ppm are achieved. Taking advantage of dual spin resonance modulation, temperature drift is suppressed to 10 ppm/K. This configuration opens new possibilities as a robust and scalable platform for current quantum sensing technologies.

physics.ins-det↗
arXiv2018

Carboxyl Carbon Quantum Dots: a Novel Type of Environmental-Friendly Scale Inhibitor

In this work, we demonstrate high inhibition efficiency of carboxyl carbon quantum dots (CCQDs) for the first time. The as-prepared CCQDs have excellent scale inhibition performance for calcium sulfate and barium sulfate. With static test of scale inhibition within temperature below 80 degrees, the anti-scaling efficiency can reach 100 % with low additions of CCQDs.

physics.app-ph↗