#multimodal large models
1548 related articles
·3 min
OpenCode Tutorial: A Complete Guide from Installation and Configuration to Hands-On Practice
Complete guide to OpenCode AI coding tool: two installation methods, model configuration, Agent types, custom commands, MCP extensions, Agent SQL, with practical examples.
Read more →
·3 min
Guide to Locally Deploying Vision Models on M4 Pro 64GB: Model Selection & Tool Recommendations
How to choose local vision language models on M4 Pro 64GB? Compare Qwen2.5-VL, Llama 3.2 Vision, and more, with tool recommendations for Ollama, LM Studio, and MLX.
Read more →
·4 min
The Rise and Fall of AI Agent Civilizations: What Multi-Agent Simulations Reveal
Explore how multi-agent simulations let AI agents autonomously build civilizations. From Stanford's AI Town to civilization-scale simulations, discover memory mechanisms, emergent behavior, and implications for social science and AI safety.
Read more →
·2 min
Ksyon: A Locally-Run AI Robot Using Moondream for Visual Perception and Lifelike Interaction
Ksyon is a fully local AI robot project using the lightweight vision-language model Moondream for environmental perception, combined with lifelike head movements and a sarcastic personality for engaging human-robot interaction.
Read more →
·1 min
Gemini Robotics 2 Deep Dive: VLA Architecture and the Future of General Embodied Intelligence
Deep dive into Google DeepMind's Gemini Robotics 2 model, analyzing its VLA architecture breakthroughs, generalization gains, dexterous manipulation advances, and the future of embodied AI.
Read more →
·2 min
Voice Agent Sandwich Architecture in Practice: STT+LLM+TTS Latency Optimization & Pitfall Guide
Deep dive into the enterprise voice agent Sandwich Architecture (STT+LLM+TTS): design principles, architecture comparison, 0.3s latency optimization, and real-world pitfalls in barge-in and tool calling.
Read more →
·3 min
Google's SKILL.state Paper Explained: Replacing History with State to Save Agents 94% of Tokens
Google's SKILL.state method replaces full conversation history with structured state, cutting Agent token usage from 1.1M to 65K (94% reduction) in 100-step benchmarks while maintaining accuracy.
Read more →
·3 min
Synthetic Tactile Datasets: Can They Break the Data Bottleneck in Robotic Grasping?
Exploring the practical value of synthetic tactile datasets for robotic grasping. Analyzing the data scarcity challenge, stick-slip physics modeling, 500K-row simulation data, and Sim-to-Real gap solutions.
Read more →
·1 min
Seven Evolutions of Retrieval Systems: The Complete Technical Lineage from BM25 to Generative Retrieval
Explore seven waves of retrieval system evolution: from BM25 lexical search to generative agentic retrieval, revealing how these technologies stack and work together in production.
Read more →
·3 min
Getting Started with AI Large Models: A Complete Technical Overview from Machine Learning to Agents
A systematic overview of the evolution from AI, machine learning, deep learning, and Transformer to LLMs, covering generative AI principles, model selection, and the future of AI Agents.
Read more →
·3 min
Cursor CEO Reveals: OpenAI Models Account for Only 5% of Traffic, Claude Becomes the Go-To for Programming
Cursor's CEO reveals OpenAI models carry only ~5% of user traffic, with 95%+ going to Anthropic Claude and competitors. A deep dive into AI coding tool model preferences and industry implications.
Read more →
·1 min
You're Probably Still Underestimating How Fast AI Models Are Evolving
Why do we always underestimate how fast AI models evolve? From linear thinking bias to exponential growth realities, and what it means for developers, investors, and users.
Read more →
·3 min
ChatGPT Voice Mode Clones User's Voice: Root Cause Analysis and Security Implications
Reddit user reports ChatGPT voice mode cloning their voice. Analysis of OpenAI's disclosed unauthorized voice generation risk, technical causes, and safety guardrail limitations.
Read more →
·3 min
OpenAI Terminates Partnership with Cursor: Model Access Cut Off November 12 — How Developers Should Respond
OpenAI will cut off model access to Cursor on November 12 following SpaceX's acquisition. Learn the reasons, developer impact, and multi-model coping strategies.
Read more →
·5 min
Tencent Hunyuan WorldClaw: Generate Editable 3D Worlds from a Single Sentence
Tencent Hunyuan's WorldClaw generates explorable, editable 3D worlds from text. Deep dive into its multi-model Agent architecture, AI-native game engines, AI pharma funding, and data strategy shifts.
Read more →
·4 min
AI Anti-Counterfeiting: Technologies and Practices for Identifying Fake Cosmetics with Artificial Intelligence
Explore how AI identifies counterfeit cosmetics through computer vision packaging inspection, spectral analysis, and multimodal detection, plus real-world challenges and blockchain-integrated anti-counterfeiting ecosystems.
Read more →
·3 min
Gemini Flash vs Pro Real-World Test: Can Being 3x Faster Really Replace It?
Real-world testing of Gemini Flash vs Pro across three projects: racing game, subscription app, and luxury website. Flash is 3x faster and cheaper, but Pro remains essential for production accuracy.
Read more →
·3 min
Alpamayo 2 Super: A Deep Dive into the Open-Source Autonomous Driving Reasoning Model
Alpamayo 2 Super is an open-source reasoning model for autonomous driving with commercial deployment support. Explore its reasoning capabilities, robotics backbone architecture, and OpenMDW-1.1 license.
Read more →
·3 min
DeepSeek V4 Pro vs. Codex: A Hands-On Comparison of AI-Recreated Don't Starve
Hands-on comparison of DeepSeek V4 Pro vs. OpenAI Codex recreating Don't Starve from scratch. DeepSeek excels at planning but gameplay breaks down; Codex delivers complete features. A deep dive into how model capability and engineering environment interact.
Read more →
·3 min
Game Behavior Data Collection: Frame-Level Aligned Human Action Data for Imitation Learning
An open-source game behavior capture tool that synchronously records gameplay video and keyboard/mouse input with frame-level alignment, providing structured datasets for imitation learning and world model research.
Read more →