Latest Articles

ByteDance Seed's New Work DeltaFormer: An Attempt at Next-Generation Model Architecture
Artificial IntelligenceNeural NetworksMachine LearningComputational ComplexityTransformer Models
...
More Toxic, More Secure? Harvard Team's Latest Research: 10% Toxic Training Makes Large Models Invulnerable
Artificial IntelligenceAI SafetyData DetoxificationMachine Learning TrainingLarge Language Models
...
LLMs Can Now Self-Update Weights, Significantly Boosting Adaptive and Knowledge Integration Capabilities. Is AI Waking Up?
Artificial IntelligenceLarge Language ModelsAI Self-ImprovementReinforcement LearningMachine Learning
...
Multi-Agent Systems Are "Burning" Tokens! Everything Anthropic Has Discovered
Artificial IntelligenceMulti-Agent SystemsSystem ArchitecturePrompt EngineeringLarge Language Models
...
Apple's 'Illusion of Thinking' Paper Criticized Again, Claude and Human Co-authored Paper Points Out Its Three Key Flaws
AI ResearchLarge Language ModelsHuman-AI CollaborationModel EvaluationAI Reasoning
...
The Meaning of the Anthropocene: Why it Matters Even Without a Formal Geological Definition | Nature Feature
AnthropoceneGeologyEarth SystemsHuman ImpactClimate Change
...
AI Acts as Its Own Network Administrator, Achieving a "Safety Aha-Moment" and Reducing Risk by 9.6%
AI SafetyLarge Language ModelsJailbreakingAI RobustnessMachine Learning Security
...
Autonomous Agent Approach is Wrong! Chinese Scholars Propose LLM-HAS: Shifting from "Autonomous Capability" to "Collaborative Intelligence"
Artificial IntelligenceLarge Language ModelsCollaborative IntelligenceAI AgentsHuman-AI Collaboration
...
Google, Amazon, Microsoft make moves: Will AI's huge energy demands revive nuclear power?
Nuclear EnergyArtificial IntelligenceEnergy TransitionSmall Modular ReactorsClean Energy
...
Amazon's New SOP Benchmark: The Ultimate Test for AI Agents. How Do Top Agents Score?
AI AgentsStandard Operating ProceduresPerformance EvaluationIndustrial AIAI Benchmarking
...
AMD and OpenAI Jointly Release Powerful AI Chips, Boosting Inference by 35x
AI HardwareAMDAI AcceleratorsData CentersOpenAI
...
The Puzzle of Free Will: Who Controls Our Choices?
NeuroscienceFree WillConsciousnessMotor ControlBrain-Machine Interface
...
Berkeley and Stanford Collaborate to Create an "AI Research Prophet": Predicting Research Idea Prospects with 77% Accuracy
AI in ResearchPredictive AIResearch MethodologyAcademic InnovationLarge Language Models
...
Kaiming He's New Work: Adding Regularization to Diffusion Models for Performance Improvement with No Pre-training or Data Augmentation, Simple to Implement
Machine LearningArtificial IntelligenceRegularizationGenerative AIDiffusion ModelsDeep Learning
...
The Economist's Reading List | 40 Best Books to Read So Far This Year
Book RecommendationsNon-fictionCurrent AffairsBusiness and TechnologyFiction
...
SRO Architecture Empowers Qwen-2.5-VL's Reasoning Capability, Boosting Performance by 16.8%
Reinforcement LearningMultimodal LLMInference ImprovementStaged OptimizationQwen-2.5-VL
...
Interview with Turing Award Laureate Joseph Sifakis: AI Can Evolve Smarter, But Cannot Fully Surpass Humans
AI EthicsAI LimitationsEducation and AISocietal Impact of AIAutonomous Systems
...
Breaking! Meta Open-Sources Its Latest World Model
Artificial IntelligenceWorld ModelsAI AgentsComputer VisionMachine Learning
...
First-Hand Review of Seedance 1.0 Pro: ByteDance's Game-Changer Dominates the Video AI Model Arena.
AI Video GenerationTechnology ReviewLarge Language ModelsArtificial IntelligenceGenerative AI
...
Altman: The Gentle Singularity Has Arrived! AI Finally Controls the Physical World, Human Destiny Takes a Major Turn in 2030
AI SingularitySam AltmanSocietal Impact of AITechnological ProgressFuture of AI
...