All Articles
16880 articles total
Gate-Zero Growth: A Geometric Framework for Function-Preserving Continual Learning
Researchers propose 'gate-zero growth,' a geometric framework for continual learning that minimizes forgetting in Transformers.
Governing Artificial Intelligence: Public Preferences and Regulatory Options
A study across seven countries reveals that citizens generally prefer safety and public governance over innovation and private regulation for AI.
Memory-Driven Self-Disclosure and Relational Turning Points: A Longitudinal Multimodal Study of Human-AI Interaction
A longitudinal study examines how memory-augmented AI agents form relational bonds with humans through self-disclosure and turning points.
Auditing Fairness-Privacy Trade-offs: Subpopulation-Level Effects of Fairness-Enhancing Algorithms
Research on the trade-offs between fairness-enhancing algorithms and privacy leakage at the subpopulation level in ML models.
Bad Memory: Evaluating Prompt Injection Risks from Memory in Agentic Systems
Evaluation of prompt injection risks in agentic systems that use persistent memory, showing that planted payloads can affect future sessions.
Angular Gaussian Supervised Contrastive Learning for Long-Tailed Electrocardiogram Arrhythmia Diagnosis
Introduction of AG-SCL, a contrastive learning framework to improve arrhythmia diagnosis in ECGs with long-tailed label distributions.
Minikotlin
Minikotlin is a project likely focused on a minimal implementation or subset of the Kotlin language.
Manufact (YC S25) Is Hiring a Senior infra engineer to build the MCP cloud
Manufact (YC S25) is hiring a Senior Infrastructure Engineer to develop their MCP cloud.
After wearing the Pebble Time 2 for two weeks, I'll never buy another smartwatch
A user shares their positive experience with the Pebble Time 2 smartwatch, suggesting it remains a superior choice compared to modern alternatives.
Even Microsoft couldn’t make Windows 11 work well on 8GB of RAM
The Verge reports that Windows 11 struggles to perform adequately on laptops with only 8GB of RAM, highlighting ongoing memory inefficiency.
Multi-Scale ViT Inference with Habitat-Fit Priors and kNN Retrieval for Multi-Species Plant Identification
Researchers developed a plant identification pipeline using DINOv2 ViT-L/14 and FAISS kNN retrieval, achieving third place in the PlantCLEF 2026 challenge.
VTM-Nav: Hierarchical Visual-Topological Memory for Cross-Episode Object-Goal Navigation
VTM-Nav introduces a hierarchical Visual-Topological Memory framework for training-free, cross-episode object-goal navigation in indoor environments.
Controlled Reformulation Testing for Logical Consistency in Large Language Models
CRTBench is a new benchmark designed to evaluate the logical consistency of LLMs across controlled reformulations of the same question.
SafeRelBench: A Spatial-Relation-Aware Benchmark for Process-Level Safety in VLM-Driven Embodied Agents
SafeRelBench is a spatial-relation-aware benchmark for evaluating process-level safety in VLM-driven embodied agents in household settings.
Answer-Conditioned Chains of Thought Degrade Verifiable-Reasoning Distillation in Large Language Models
Research shows that distilling reasoning from LLMs using answer-conditioned chains of thought leads to 'backward rationalization' and degrades verifiable reasoning accuracy.
A Modern Multimodal Assistant on a 6 GB 2011 GPU: Stage-Validated, All-GPU CUDA Inference for Fermi
A technical report demonstrates running the MiniCPM-V-4.6 multimodal assistant entirely on a 2011 NVIDIA Fermi GPU through optimized CUDA kernels and 8-bit quantization.
Apple targets dozens of OpenAI employees with legal letters
Apple is reportedly sending legal letters to former employees who joined OpenAI, potentially signaling a legal battle over talent acquisition and trade secrets.
Trump is selling high-speed access to his market-moving Truth Social posts
Trump Media is launching a paid API for Truth Social to provide Wall Street firms with fast access to market-moving posts.
Decision Making Needs Uncertainty Quantification [Lecture Notes]
A set of lecture notes exploring the theoretical link between decision-making objectives, uncertainty representation, and optimal agent performance.
ConFlow: Constraints-Guided Learning with Flow Matching for Motion Generation
ConFlow introduces a constraint-guided flow matching framework for robot motion generation that integrates constraints directly into the training objective.