NORACL: Solving the Stability-Plasticity Dilemma in Continual Learning with Neurogenesis Mechanisms
A latest arXiv paper proposes the NORACL framework, drawing inspiration from biological neurogenesis mechanisms to achie…
📂 Research
Latest articles in Research
How Dynamic Adversarial Fine-Tuning Reshapes Model Refusal Geometry
A latest arXiv study reveals that dynamic adversarial fine-tuning reorganizes the refusal directions of language models …
The Continual Learning Dilemma of Memory-Augmented LLM Agents Re-Exposed
New research reveals that memory-based LLM agents have not truly solved the stability-plasticity dilemma in continual le…
Topological Methods Achieve Breakthrough in Monitoring Neural Network Training Collapse
A latest arXiv paper proposes an online topology-aware monitoring method based on Modular Morse Homology Maintenance (MM…
Masked Diffusion Models Get a Clean Upgrade: Self-Conditioning Adaptation Significantly Boosts Generation Quality
A latest arXiv paper proposes a simple self-conditioning adaptation method for Masked Diffusion Models (MDMs). By reusin…
FairMind: Automating AI Fairness Analysis with Causal Reasoning and LLMs
A research team has introduced FairMind, a prototype system that combines causal fairness analysis with LLM-powered repo…
New Study Proposes Multi-Dimensional Framework for Evaluating LLM-Generated Clinical Synthetic Data
A new study introduces a multi-dimensional evaluation framework assessing LLMs including DeepSeek-R1 across fidelity, di…
People First: A Paradigm Shift in Medical Imaging AI
A new study reveals that while data-driven medical AI has achieved high diagnostic accuracy, its clinical adoption remai…
Digital Twin + AI: A New Framework for Personalized Cognitive Decline Assessment
A research team has proposed the PCD-DT framework, which integrates multimodal data with uncertainty modeling to build p…
New Study Reveals the 'Consensus Paradox' in AI Multi-Agent Swarms
A new study based on 36 experiments and 12,804 decision trajectories challenges the fundamental 'wisdom of the crowd' as…
OptimusKG: A Multimodal Knowledge Graph Unifying Biomedical Knowledge
A research team has released OptimusKG, a multimodal biomedical knowledge graph based on labeled property graphs, design…
AutoSurfer: Comprehensive Surf-Style Training Makes Web Agents Stronger
A latest arXiv paper proposes the AutoSurfer framework, which addresses the scarcity of training data and incomplete web…