Q: Why choose CISPO instead of GSPO or GRPO? How well does CISPO adapt to MoE, and does changing the RL algorithm require architectural refactoring? GRPO predates both, but in our attempts to reproduce R1-Zero it proved unreliable: PPO-style clipping caused token-level gradients

Kimi K2.5 tech report just dropped! Quick hits: - Joint text–vision training: pretrained with 15T vision-text tokens, zero-vision SFT (text-only) to activate visual reasoning - Agent Swarm + PARL: dynamically orchestrated parallel sub-agents, up to 4.5× lower latency, 78.4% on

GLM 4.7 is one of the strongest open-source coding models available—but most developers aren't prompting it correctly. We put together 10 rules to help you get the most out of it: - Front-load instructions (it has a strong recency bias) - Use firm language: "must" and

Self-Distillation Enables Continual Learning MIT & ETH Zurich researchers introduce SDFT for on-policy learning from demonstrations. It uses demonstration-conditioned models as their own teacher to reduce catastrophic forgetting, outperforming SFT and enabling sequential skill

CTF as a Benchmark

挖国内云厂商漏洞前真心建议多刷刷wiz的题，还有云鼎早期的文章，80万月榜不是梦😄。很多同学都没掌握挖云厂商漏洞精髓，主动放弃了。还有一些被云产品的价格劝退了，说实话舍不得孩子套不着狼😑….

Qwen-Coder-Qoder is live! We’ve launched a customized model built on Alibaba’s Qwen-Coder, fine-tuned via large-scale RL specifically for the Qoder. Learn more: qoder.com/blog/qwen-code…

Qwen releases Qwen3-Coder-Next. 💜 The new 80B MoE model excels at agentic coding & local use. With 256K context, it delivers similar performance to models with 10-20× more active parameters. Run on 46GB RAM or less. Guide: unsloth.ai/docs/models/qw… GGUF: huggingface.co/unsloth/Qwen3-…

JUST IN: Ethereum Founder Vitalik Buterin sold 2,972 ETH for $6.69 million over the past 3 days.

【合格体験記】カーネルモードから見る防御 ― CETP合格までの軌跡 gmo-cybersecurity.com/blog/evasion-l…

you can use gmod to verify your age by the way

We're thrilled to lead Shizuku AI's seed round. In January 2023, while still completing his Ph.D. at UC Berkeley, Akio Kodaira launched an AI VTuber named Shizuku on YouTube. He ran dozens of streams, building a community of thousands of followers who tuned in to talk with an AI

当社エンジニアの金子 孟司が発見した攻撃手法がPortSwigger社の「Top 10 Web Hacking Techniques of 2025」に選出されました gmo-cybersecurity.com/news/20260210/

Introducing GLM-5: From Vibe Coding to Agentic Engineering GLM-5 is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5, it scales from 355B params (32B active) to 744B (40B active), with pre-training data growing from 23T to 28.5T tokens.

GLM なのに DeepSeek Sparse Attention

🚀 Unveiling Ring-1T-2.5 The first hybrid linear-architecture 1T thinking model. -Efficient: Hybrid linear breakthrough (10x lower memory) -Gold Tier: IMO25 (35/42) & CMO25 (105/126) -Agentic: Natively with Claude Code & OpenClaw -Open SOTA: IMOAnswerBench，GAIA2-search & more!

Introducing M2.5, an open-source frontier model designed for real-world productivity. - SOTA performance at coding (SWE-Bench Verified 80.2%), search (BrowseComp 76.3%), agentic tool-calling (BFCL 76.8%) & office work. - Optimized for efficient execution, 37% faster at complex

We built a Complete Replacement Model (CRM) that fully sparsifies a language model. This brings many changes to circuit tracing and global circuits. (1/n)

BQTLockとGREENBLOODについてANYRUNより Go製のランサムウェアもある。。。書きやすそうではある。ちゃんと侵入後の検知手法も書いてあるいい記事！ any.run/cybersecurity-…

投資家にとっても起業家にとっても IPO は夢