Thrilled by the interest in our work! [Surprising finding] Across 7,650 source-target pairs, neural networks with LLM embeddings outperformed GBDTs, DROs, and GPT-4-mini. See the 'optimal ratio' for each method below 👇 [Question] Do we still need a dedicated Tabular LLM? 🤔

Probably the 1ST data-centric LLM copilot, integrating various data-centric tools, supporting complex data processing and modeling of clinical problems🔥🔥 - Paper: Summarizes various data-related issues & corresponding tools, along with detailed case studies in the medical field

Thank you so much! YES, we find Grok 4 super cool, and it is really good at harder tiers (Deep Search and Super Agent). Btw, our case study (sec 4.5.1) shows that Grok 4’s search capability is much better!

Shout out to our great advisors Mengdi Wang Jose H. Blanchet 邱锡鹏 Rubio Huang and our ByteDance team!

We're collaborating with six commercial agents to include them in the next leaderboard update. Meanwhile, GPT-5 Pro, ChatGPT Agent, Gemini 2.5 Pro Deep Think, and Claude Opus 4.1 are all on the way. And now, a future prediction challenge: Will Grok remain the best? Let’s see…

Wow thanks Elon! Yes, we think it's a measure of agent's AGI! Again, introduce our FutureX live benchmark. futurex-ai.github.io