arXiv subjects

Jianxiong Zhou

Publications and source records attributed to Jianxiong Zhou.

3 recordsLinked to original sources

arXiv2026

NeuroWorld: A Latent Brain World Model for Stimulus-Conditioned Human Brain Dynamics

Forecasting human brain activity during naturalistic experience requires modeling how endogenous neural states evolve causally under continuous sensory drive. Existing brain encoding models instead frame this as stimulus-to-response regression without strict temporal constraints, allowing future stimuli to leak into current predictions. We introduce NeuroWorld, to our knowledge the first brain world model, which casts naturalistic brain functional dynamics prediction as stimulus-conditioned evolution in a learned latent brain-state space, separating endogenous states (measured via fMRI) from exogenous multimodal stimuli across two stages. Latent Dynamics Learning (LDL) jointly learns a transition-sufficient representation and causal dynamics through next-latent prediction, without reconstructing the observed fMRI signal. Latent Rollout Decoding (LRD) freezes LDL, autoregressively rolls latent states forward from an observed fMRI prefix, and decodes them into subject-specific whole-brain responses. Across three naturalistic movie-fMRI benchmarks spanning 30 participants, including our newly collected Singapore Multimodal Imaging & Naturalistic Dataset (SG-MIND; 20 participants, 8,519 paired stimulus-response clips, 140.7 person-hours of viewing), NeuroWorld achieves state-of-the-art multi-step rollout performance under strictly causal stimulus access, with greater robustness to long-horizon autoregressive drift, supporting reliable simulation of extended brain-state trajectories. Extensive interpretability analyses characterize the functional organization of the learned dynamics, establishing latent-space world modeling as a principled framework for causal forecasting of human brain activity.

q-bio.NC↗

arXiv2026

BrainFIBRE: A Foundation Model via Information Decomposition for Brain Microstructure

Diffusion MRI probes brain microstructure with particular sensitivity to early cerebrovascular and neurodegenerative changes. Neurite Orientation Dispersion and Density Imaging (NODDI) decomposes the diffusion signal into three biophysically interpretable maps: neurite density index (NDI), orientation dispersion index (ODI), and free water fraction (FWF), capturing neurite packing, fiber coherence, and extracellular fluid. These 3D maps offer a rich substrate for transferable microstructural representations, yet integrating them is challenging: standard representation learning struggles to disentangle the unique information in each map from their shared and synergistic interactions. We present BrainFIBRE, the first foundation model for brain microstructure, pretrained on NODDI-derived maps from 55,592 UK Biobank participants. We propose Self-supervised Partial Information Decomposition (SPID), which extends PID-guided multimodal learning to the self-supervised regime for the first time. A novel Counterfactual Candidate Construction (CCC) paradigm perturbs inter-modality alignment through modality dropping and swapping, providing the contrastive signal for a Mixture-of-Experts architecture to disentangle unique, synergistic, and redundant information without any downstream label. On both Caucasian and Asian cohorts, BrainFIBRE achieves state-of-the-art performance across diverse tasks predicting age, sex, cerebrovascular and neurodegenerative markers, and cognition, while yielding neurobiologically interpretable representations that reveal task- and cohort-specific interaction patterns. BrainFIBRE establishes a versatile foundation for neuroimaging analysis at the microstructural level.

cs.CV↗

arXiv2023

Active Open-Vocabulary Recognition: Let Intelligent Moving Mitigate CLIP Limitations

Active recognition, which allows intelligent agents to explore observations for better recognition performance, serves as a prerequisite for various embodied AI tasks, such as grasping, navigation and room arrangements. Given the evolving environment and the multitude of object classes, it is impractical to include all possible classes during the training stage. In this paper, we aim at advancing active open-vocabulary recognition, empowering embodied agents to actively perceive and classify arbitrary objects. However, directly adopting recent open-vocabulary classification models, like Contrastive Language Image Pretraining (CLIP), poses its unique challenges. Specifically, we observe that CLIP's performance is heavily affected by the viewpoint and occlusions, compromising its reliability in unconstrained embodied perception scenarios. Further, the sequential nature of observations in agent-environment interactions necessitates an effective method for integrating features that maintains discriminative strength for open-vocabulary classification. To address these issues, we introduce a novel agent for active open-vocabulary recognition. The proposed method leverages inter-frame and inter-concept similarities to navigate agent movements and to fuse features, without relying on class-specific knowledge. Compared to baseline CLIP model with 29.6% accuracy on ShapeNet dataset, the proposed agent could achieve 53.3% accuracy for open-vocabulary recognition, without any fine-tuning to the equipped CLIP model. Additional experiments conducted with the Habitat simulator further affirm the efficacy of our method.

cs.CV↗