Latest

Two RTX 4090s Can Locally Fine-Tune the Trillion-Parameter Kimi K2 — Qujing, Tsinghua, and BUAA Smash the Computing Power Barrier

RTX 4090

Two RTX 4090s Can Locally Fine-Tune the Trillion-Parameter Kimi K2 — Qujing, Tsinghua, and BUAA Smash the Computing Power Barrier

# Fine-Tuning Ultra-Large Parameter Models — Now Possible on Consumer GPUs Fine-tuning extremely large models has undergone a **dramatic transformation**: You now only need **2–4 consumer-grade GPUs** (such as the RTX 4090) to locally fine-tune models as huge as **DeepSeek 671B** or **Kimi K2 1TB**. ![image](https://blog.aitoearn.ai/content/

By Honghao Wang
RedCodeAgent Red Team Assessment for Improving Code Agent Security Risks

RedCodeAgent

RedCodeAgent Red Team Assessment for Improving Code Agent Security Risks

Introduction Code agents are AI-powered systems designed to generate high-quality code and work seamlessly with code interpreters. They streamline complex software development workflows, which has driven widespread adoption. However, these benefits come with significant safety and security challenges. Existing static safety benchmarks and traditional red-teaming approaches—where security researchers simulate

By Honghao Wang
How to Make Agents Meet Expectations: Top 10 Practical Lessons from Building Cloud Assistant Aivis with Context Engineering and Multi-Agent Systems

AI agents

How to Make Agents Meet Expectations: Top 10 Practical Lessons from Building Cloud Assistant Aivis with Context Engineering and Multi-Agent Systems

Building & Optimizing AI Agents: Lessons from YunXiaoer Aivis This is the 123rd article of 2025 (Estimated reading time: 15 minutes) --- 01. Background This year, our team has been focusing on the YunXiaoer Aivis project — a digital employee in the Alibaba Cloud service domain. It represents our transition from

By Honghao Wang
Better Chinese Skills and Finer Control Than NanoBanana! Tuzhan & Peking University Uniworld V2 Sets New SOTA

Better Chinese Skills and Finer Control Than NanoBanana! Tuzhan & Peking University Uniworld V2 Sets New SOTA

# UniWorld-V2: Next-Level Precision Image Editing with Reinforcement Learning A **new model**, even better than Google’s Nano Banana at **fine-detail image editing** and with **stronger Chinese understanding**, has arrived — introducing **UniWorld-V2**. --- ## 📌 Quick Comparison Example **Task:** “Change the hand gesture of the girl in the middle wearing a white shirt

By Honghao Wang