Lun Wang (@lunwang1996) 's Twitter Profile
Lun Wang

@lunwang1996

Senior Research Scientist @GoogleDeepMind on Gemini Live. PhD @UCBerkeley. Multimodal LLM & LLM safety. Fishing enthusiast. Opinions are my own.

ID: 871176931227516928

linkhttps://wanglun1996.github.io calendar_today04-06-2017 01:29:00

96 Tweet

1,1K Followers

111 Following

John Lambert (@jlambert_) 's Twitter Profile Photo

I am hiring a Student Researcher at Google DeepMind for 2025! 👩🔬 Interested in improving multi-turn optimization and reasoning capabilities of LLMs? 🧑‍🎓 Currently studying for a Bachelor's/Master's/PhD? 🧑‍💻 Have solid engineering and research skills? 🌟We want to hear from you!

I am hiring a Student Researcher at Google DeepMind for 2025!

👩🔬 Interested in improving multi-turn optimization and reasoning capabilities of LLMs?
🧑‍🎓 Currently studying for a Bachelor's/Master's/PhD?
🧑‍💻 Have solid engineering and research skills?

🌟We want to hear from you!
Lun Wang (@lunwang1996) 's Twitter Profile Photo

Short-term, utility > safety >> privacy > explainability. But long-term? Explainability is needed to ASI. We’re still in AI’s alchemy phase—waiting for our Lavoisier.

OpenAI (@openai) 's Twitter Profile Photo

Detecting misbehavior in frontier reasoning models Chain-of-thought (CoT) reasoning models “think” in natural language understandable by humans. Monitoring their “thinking” has allowed us to detect misbehavior such as subverting tests in coding tasks, deceiving users, or giving

Detecting misbehavior in frontier reasoning models

Chain-of-thought (CoT) reasoning models “think” in natural language understandable by humans. Monitoring their “thinking” has allowed us to detect misbehavior such as subverting tests in coding tasks, deceiving users, or giving
Yi Tay (@yitayml) 's Twitter Profile Photo

The patent process is hilariously slow 🐢. Just found out that my first patent with Google was just approved. The process was so slow that in this time frame I wrote the paper for it, got it accepted, moved countries, got married, left google, had a kid, did a startup, left the

Google DeepMind (@googledeepmind) 's Twitter Profile Photo

Think you know Gemini? 🤔 Think again. Meet Gemini 2.5: our most intelligent model 💡 The first release is Pro Experimental, which is state-of-the-art across many benchmarks - meaning it can handle complex problems and give more accurate responses. Try it now →

Peyman Milanfar (@docmilanfar) 's Twitter Profile Photo

You should be so lucky to have people throughout your research career that you can openly bounce ideas to and from - especially if they complement your strengths in your areas of weakness - it is a rare and precious gift.

You should be so lucky to have people throughout your research career that you can openly bounce ideas to and from - especially if they complement your strengths in your areas of weakness - it is a rare and precious gift.
Xin Eric Wang @ ICLR 2025 (@xwang_lk) 's Twitter Profile Photo

Engineering matters more than ever in AI. If you’re working on core LLM training or agent building and dismiss the engineering effort, it’s likely you’re not a core contributor.

Lun Wang (@lunwang1996) 's Twitter Profile Photo

I'm moving all my crypto off a major exchange (I believe you will know which if you trade crypto) because my account is suspended. While I still appreciate it allowing me to transfer to other wallets, their support is a complete mess. I still don't understand the reason for the

Lun Wang (@lunwang1996) 's Twitter Profile Photo

The key contribution: we show it's possible to craft noise that survives being played through the air—not just digital input manipulation. That makes this a much more realistic threat: Imagine someone playing noisy music near you to trigger the LLM agent on your phone to change

The key contribution: we show it's possible to craft noise that survives being played through the air—not just digital input manipulation. That makes this a much more realistic threat: Imagine someone playing noisy music near you to trigger the LLM agent on your phone to change
Google DeepMind (@googledeepmind) 's Twitter Profile Photo

An advanced version of Gemini with Deep Think has officially achieved gold medal-level performance at the International Mathematical Olympiad. 🥇 It solved 5️⃣ out of 6️⃣ exceptionally difficult problems, involving algebra, combinatorics, geometry and number theory. Here’s how 🧵

An advanced version of Gemini with Deep Think has officially achieved gold medal-level performance at the International Mathematical Olympiad. 🥇

It solved 5️⃣ out of 6️⃣ exceptionally difficult problems, involving algebra, combinatorics, geometry and number theory. Here’s how 🧵