Zihan "Zenus" Wang (@wzenus) / X

Zihan "Zenus" Wang

1,237 posts

Zihan "Zenus" Wang

@wzenus

Reasoning agent / RL / efficiency research @NorthwesternU & @nvidia. Ex @Microsoft @yutori_ai @deepseek_ai @uiuc_nlp @RUC1937.

Joined March 2022

Pinned
Zihan "Zenus" Wang
@wzenus
May 29
🧵 Claude-Opus-4.8 takes you too much tokens - but is this issue general across agents? Do agents know how much they'll spend? Introducing Budget-Aware Agents (BAGEN): We study budget awareness across 4 envs & 5 frontier agents, and find structured failures in most of them.
607K
Zihan "Zenus" Wang
@wzenus
Jan 28, 2025
Wow. So why didn't you open-source it?
Mark Chen
@markchen90
Jan 28, 2025
Congrats to DeepSeek on producing an o1-level reasoning model! Their research paper demonstrates that they’ve independently found some of the core ideas that we did on our way to o1.
3.1M
Zihan "Zenus" Wang
@wzenus
Jan 24, 2025
DEEPSEEK NOW IS THE #1 IN THE WORLD. 🌍🚀 Never been prouder to say I got to work here. Ambition. Grit. Integrity. That’s how you build greatness. Brilliant researchers, engineers, all-knowing architects, and visionary leadership—this is just the beginning. Let’s. Go. 💥🔥
Arena.ai
@arena
Jan 24, 2025
Breaking News: DeepSeek-R1 surges to the top-3 in Arena🐳! Now ranked #3 Overall, matching the top reasoning model, o1, while being 20x cheaper and open-weight! Highlights: - #1 in technical domains: Hard Prompts, Coding, Math - Joint #1 under Style Control - MIT-licensed A
1M
Zihan "Zenus" Wang
@wzenus
Jan 28, 2025
Not responding to more interview request anymore (if you do think there is a need, email me tho). The best way to ask me questions is through X, where all people from the world could see who is correct who is not. Being wrong is not bad, being arrogant is.
623K
Zihan "Zenus" Wang
@wzenus
Jan 28, 2025
I don't think Anthropic is behind. Please give them (as well as other respectful research teams) more time. There is an old saying in China: if you become more low-key, you will become richer (中国有句古话叫做，闷声发大财。). How they are coping with non-coping when others are
Theo - t3.gg
@theo
Jan 28, 2025
Prediction: Anthropic will rush out a reasoning model within the next 30 days
339K
Zihan "Zenus" Wang
@wzenus
Jan 28, 2025
🚀 Introducing RAGEN—the world’s first reproduction of DeepSeek-R1(-Zero) methods for training agentic AI models! We’re betting big on the future of RL + LLM + Agents 🤖✨. This release is a minimally viable leap toward that vision. Code and more intro 🔗:
356K
Zihan "Zenus" Wang
@wzenus
Dec 28, 2024
[Long Tweet Ahead] I just have to say, I’m genuinely impressed by DeepSeek. 💡 It’s no wonder their reports are so elegant and fluffless. Here’s what I noticed about their culture, a space where real innovation thrives, during my time there ↓ — — — — — 🌟 1. Be nice and careful
266K
Zihan "Zenus" Wang
@wzenus
Jan 28, 2025
Replying to @wzenus
I have no offense here (I know you don't believe). I'm just wondering how they are becoming more and more tentative about what has been their advantage and the most proud personality of the "Open AI" company, focusing on what's really important, and trust people like how people
175K
Zihan "Zenus" Wang
@wzenus
Jan 24, 2025
Things I'm worrying now is how "tech old moneys" are trying to offer DeepSeek whales 🐳 those 💵5.5M/yr salary, hoping to dissolve the team and disrupt such opponent. No. I'd never want this to happen, and whale HRs please keep offering your members the best material benefits
97K
Zihan "Zenus" Wang
@wzenus
Jan 28, 2025
Replying to @wzenus
These days DeepSeek has been gathering more and more attention. While I'm not sure if I have the right, I think it's a great time to ask some questions to someone that worth being asked. It's not about competition or war, it's about the future of humanity. Do not be afraid of
172K
Zihan "Zenus" Wang
@wzenus
Mar 3, 2025
🚀 Introducing Chain-of-Experts (CoE), A Free-lunch optimization method for DeepSeek-like MoE models! within $200, we explore to train MoEs that enables 17.6-42% efficiency boost in memory! Code: github.com/ZihanWang314/c… Blog: notion.so/Chain-of-Exper…
135K
Zihan "Zenus" Wang
@wzenus
Dec 25, 2024
Rare life update: I'm married! 💍 And so surprised to hit "521" ("I love you" in Chinese) citations the same day :）It's really a great time and start of a new journey. Merry Christmas and happy holidays to everyone!🎄🎉🎅🏻
86K
Zihan "Zenus" Wang
@wzenus
Aug 6, 2025
To guys diving into fine-tuning open-source MoEs today: check out ESFT, our customized PEFT method for MoE models. Train with 90% less parameters, gain 95%+ task perf and keep 98% general perf :)
DeepSeek
@deepseek_ai
Jul 5, 2024
🚀 Introducing Expert-Specialized Fine-Tuning (ESFT) for Customizing LLMs with Sparse Architectures! 🌟 Highlights: - Train only task-relevant experts for LLM customization. - Reduces storage by up to 90% and training time by up to 30%. ✨ Performance: - Customizes LLMs
48K
Zihan "Zenus" Wang
@wzenus
Jan 28, 2025
If you'd ever been inspired by @deepseek_ai , Don't miss following @teortaxesTex — DeepSeek's Cheerleader since day 1, as early as 2023, back when ZERO followers roamed here 🦗 Never underestimate this cheering maestro 💥 His words? Unpredictable. His takes? Fresh. His energy?
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
@teortaxesTex
Jan 27, 2025
Badge of honor if I've ever had one
102K