Topic
representation learning
New Framework GeoVR Learns 3D Spatial Intelligence from 2D Videos for Multimodal LLMs
Multimodal Large Language Models (MLLMs) traditionally lack intrinsic 3D awareness. Researchers present GeoVR, a framework that learns geometric representations from 2D video sequences, restructuring the semantic latent space to unlock spatial intelligence. GeoVR uses four complementary geometric targets from pre-trained 3D foundation models, achieving state-of-the-art performance on spatial reasoning benchmarks.
REST-GAN: A Deep Generative Model for Resting-State EEG Synthesis and Transferable Representation Learning
Researchers introduce REST-GAN, a generative adversarial network for resting-state EEG that both synthesizes realistic neural signals and learns transferable representations. The model achieves high precision and recall in band-power features and shows competitive performance in demographic classification tasks, requiring substantially less training data and computational resources than existing methods.
Boundary Embedding Shaping with Adaptive Contrastive Learning Boosts GNN Classification by 3.3%
Graph neural networks suffer from structural entanglement, especially near class boundaries. A new plug-in module called Boundary Embedding Shaping (BES) uses adaptive contrastive learning to selectively suppress spurious correlations, boosting GCN node classification by an average of 3.3% (up to 5% on WikiCS) and improving link prediction accuracy.
Concept Flow Models Anchor AI Reasoning with Hierarchical Bottlenecks to Reduce Information Leakage
Researchers Wang and Paschke propose Concept Flow Models (CFMs) that replace the flat bottleneck in Concept Bottleneck Models (CBMs) with a hierarchical, concept-driven decision tree. CFMs mitigate information leakage by reducing effective concept usage, matching predictive performance of flat CBMs while providing stepwise decision flows for transparent and auditable model reasoning.
FOUNDv2: Unified Quantized Tokenizers Transform User Representation Learning
FOUNDv2, a novel user representation framework, uses quantized tokenizers to transform heterogeneous data into discrete tokens, achieving superior performance and reduced storage costs. The model outperforms task-specific baselines and has been deployed at scale on Alipay, demonstrating practical efficiency.
Representation Autoencoders v2 Achieves 10x Faster Convergence and State-of-the-Art Image Generation
A team of researchers has introduced RAEv2, an improved version of Representation Autoencoders (RAE), which achieves state-of-the-art image generation results with over 10x faster convergence. The work reveals that RAE and representation alignment (REPA) are complementary, and that REPA can provide guidance for classifier-free diffusion without a second weaker model.
Cascaded Sparse Autoencoders Enable Hierarchical Visual Concept Learning in Multimodal LLMs
Researchers introduce cascaded sparse autoencoders (CSAEs) that learn hierarchical visual concepts in multimodal large language models. By training a second-level SAE on the decoder weights of the first, CSAEs achieve 'concepts of concepts' without nesting or stacking bottlenecks. Experiments on Qwen3-VL, Gemma-3, and LLaVA show improved interpretability and effective group-level steering.
Autonomous End-to-End SOH Prediction Service Uses Temporal-Contrastive Learning to Cut Error by Half
A new plug-and-play service architecture called TC-SOH uses temporal-contrastive representation learning to predict lithium-ion battery state of health directly from raw operational data, eliminating manual feature engineering. Across four public datasets, it reduces mean absolute percentage error by 1.91 times and root mean squared error by 2.13 times compared to physics-informed and data-driven baselines. The approach also improves model transparency through a suite of representation diagnostics, including visualization and sensitivity analysis.
Sensor-Conditioned Representation Learning Uses Scene-Relevant Observation Quotients to Improve Latent Geometry
Researchers propose a sensor-conditioned representation learning framework using scene-relevant observation quotients. Their OQ-TSAE method, tested on synthetic and real-radar data, improves representation-correctness diagnostics over reconstruction, metric-learning, and contrastive baselines.
RECTOR Framework Sets New State-of-the-Art in EEG Emotion Recognition and sEEG Classification
Researchers propose RECTOR, a self-supervised framework for representation learning from EEG/sEEG data, achieving state-of-the-art performance in emotion recognition and task-engagement classification. The model demonstrates strong robustness to missing channels and cross-montage generalization, promising for large-scale pre-training on heterogeneous neural data.
EyeMVP AI Model Enhances Retinal Screening by Learning OCT Insights from Fundus Photos
Researchers developed EyeMVP, a cross-modal retinal foundation model that enriches color fundus photography (CFP) with depth-resolved information from optical coherence tomography (OCT). Pretrained on 674,893 paired images from 112,642 patients across eight Chinese hospitals, EyeMVP outperforms leading models on 16 downstream tasks including macular edema detection (AUROC 0.948 vs 0.852) and myopic macular schisis (0.825).
New Rational Sparse Autoencoder Improves AI Interpretability with Trainable Activation Function
Researchers introduce the Rational Sparse Autoencoder (RSAE), which replaces fixed encoder nonlinearities with a trainable rational function. Across three language models and three baseline activation families, RSAE strictly improves reconstruction and downstream-behaviour metrics while preserving feature-level interpretability, adding only a few scalar parameters per autoencoder.
Subject-Specific Encoders Improve Cross-Subject EEG Decoding, Study Finds
A new study on arXiv.org proposes replacing shared EEG encoders with subject-specific encoders to handle inter-subject distribution shifts. The hybrid model, tested on four motor-imagery datasets, internalises Euclidean Alignment and increases class distinctiveness, though head selection for unseen subjects remains a bottleneck.