Arian Hosseini

Research Scientist @GoogleDeepMind - LLM post-training - prev: @Google @MSFTResearch

[Arian Hosseini @arianTBD]

Our team at Google DeepMind is hiring a Research Scientist in Mountain View!
Focus is on bridging foundational research & practical GenAI deployments.
✅ PhD in CS/ML
✅ LLMs & diffusion expertise
✅ Top-tier ML pubs
✅ Python/JAX/PyTorch
Apply : https://t.co/paNhfixL8l

[Arian Hosseini @arianTBD]

@marcgbellemare @reliant_ai @cohere Congratulations 🎉

[Arian Hosseini @arianTBD]

@murefil verification is still 🔑

[Arian Hosseini @arianTBD]

@BlackHC Thank you 🙏🏻

[Arian Hosseini @arianTBD]

Our team at GDM is hiring a Student Researcher (SR) next year 🧠

If you’re a PhD student working on LLMs please apply. I’d love to hear from you.

Please fill out this form: https://t.co/BVBVb5kXhA

[Arian Hosseini @arianTBD]

The new Gemini is out and the reasoning capabilities are next level 🧠 fun one to be a part of. Congrats to the whole team!

[Arian Hosseini @arianTBD]

@Mila_Quebec @sivareddyg @CSCan_InfoCan Big congratulations @sivareddyg 🎊

[Arian Hosseini @arianTBD]

Come by ScaleR #COLM2025. Nishad and I will be presenting this work!

[Nishad Singhi (at CVPR) @nishadsinghi]

Want to know how to optimally scale test-time compute with generative verifiers?
Come to our poster at #COLM2025 🇨🇦.

hmu if you want to chat about LLM reasoning and agents! :)

Poster session 1 (Tuesday Morning 11 AM - 1 PM), Poster number: 24.

[Milad Aghajohari @MAghajohari]

Introducing linear scaling of reasoning:
𝐓𝐡𝐞 𝐌𝐚𝐫𝐤𝐨𝐯𝐢𝐚𝐧 𝐓𝐡𝐢𝐧𝐤𝐞𝐫

Reformulate RL so thinking scales 𝐎(𝐧) 𝐜𝐨𝐦𝐩𝐮𝐭𝐞, not O(n^2), with O(1) 𝐦𝐞𝐦𝐨𝐫𝐲, architecture-agnostic.

Train R1-1.5B into a markovian thinker with 96K thought budget, ~2X accuracy 🧵

[Arian Hosseini @arianTBD]

Come by # 82 at #COLM2025

[Kusha Sareen @KushaSareen]

Presenting this tomorrow at COLM :) Come say hello @ 4:30 #82

[Arian Hosseini @arianTBD]

@nishadsinghi is presenting our work on compute optimal test-time scaling at @COLM_conf 🇨🇦

11am-1pm poster # 24

[Nishad Singhi (at CVPR) @nishadsinghi]

Want to know how to optimally scale test-time compute with generative verifiers?
Come to our poster at #COLM2025 🇨🇦.

hmu if you want to chat about LLM reasoning and agents! :)

Poster session 1 (Tuesday Morning 11 AM - 1 PM), Poster number: 24.

[Arian Hosseini @arianTBD]

@LiamFedus @ekindogus @periodiclabs Huge congratulations @LiamFedus and @ekindogus

RL from nature feedback sounds super cool

[Arian Hosseini @arianTBD]

@arkitus Taarof is an emergent behavior

[Arian Hosseini @arianTBD]

@M_R_Samsami @GoogleDeepMind Congrats Mo 🎊

[Arian Hosseini @arianTBD]

@hugo_larochelle @Mila_Quebec Congratulations Hugo 🎊 Mila is in good hands

[Mehran Kazemi @kazemi_sm]

Our new work probes LLMs’ multi-turn reasoning and planning capabalities via five challenging games. Check it out!

[Arian Hosseini @arianTBD]

MTP provides a robust platform to diagnose weaknesses and improve multi-turn capabilities.

Enjoyed working with @BadolaKartikeya @kazemi_sm @BahareFatemi Jon Simon and the rest of the team @GoogleDeepMind .

Details: https://t.co/iSG4HtyBSY
Data & more: https://t.co/2pHMwjyQki

[Arian Hosseini @arianTBD]

LLMs are great at single-shot problems, but in the era of experience, interactive environments are key 🔑

Introducing * Multi-Turn Puzzles (MTP) * , a new benchmark to test multi-turn reasoning and strategizing

🔗 Paper: https://t.co/iSG4HtyBSY
🫙Data: https://t.co/0bRlKfqp5W

[Arian Hosseini @arianTBD]

Our analysis shows stronger models don't just guess better. they build coherent plans, ask complex questions, and re-evaluate their own conclusions based on new info.

Weaker models jump to conclusions and ask less or questions. It's simple Q&A vs strategic dialogue.

7/n