Mark Ibrahim

Researching the dark arts of deep learning at FAIR, Meta Superintelligence Labs

[Dr. Karen Ullrich @karen_ullrich]

@srush_nlp @kchonyc @ylecun @sleepinyourhat @togelius @andrewgwils @Andrew__Brown__ @LerrelPinto @tallinzen @stoyanoj @thegautamkamath @KempeLab @cvondrick @yoavartzi

[Mark Ibrahim @marksibrahim]

New work from our team, led by @yuvalkansal : teaching agents to learn from their mistakes.

Agents are just as likely to repeat a mistake in a new session. We propose Self-Direction so agents learn from experience by updating their own weights.

[Yuval Kansal @yuvalkansal]

Agents can often fix a mistake by retrying, then forget what they just learned! Next session, same mistake 🥲

Introducing Self-Direction: each failure becomes a small, procedural adapter the agent trains on its own reflection and reuses later 💪🏽

🌐 https://t.co/SCCrlSf5ZT
🧵1/n

[Yuval Kansal @yuvalkansal]

Agents can often fix a mistake by retrying, then forget what they just learned! Next session, same mistake 🥲

Introducing Self-Direction: each failure becomes a small, procedural adapter the agent trains on its own reflection and reuses later 💪🏽

🌐 https://t.co/SCCrlSf5ZT
🧵1/n

[Mark Ibrahim @marksibrahim]

@candacerossio @eshikasax @adinamwilliams + @arreqe_ai

[Mark Ibrahim @marksibrahim]

Muse Spark Open, an open-weight frontier model that can run on two of the latest Mac Studios: https://t.co/F8QB8xIO8H

Proud of our team’s contributions!
@candacerossio @eshikasax @adinamwilliams + many others

[AI at Meta @AIatMeta]

Today we’re releasing Muse Spark Open, the most capable US open-weight model to date.

Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.

Key Model Specs:
→ 816B total / 42B active parameters
→ 1M token context window
→ Natively multimodal across text, image, video, audio

[Reyhane Askari @ReyhaneAskari]

Don’t miss @dohmatobelvis presenting our latest work, “Why less is more (sometimes): A theory of data curation” at #ICLR2026!

​Swing by our poster at the main conference to chat:
📅 Saturday, April 25
🕒 3:15pm–5:45pm
📍 Pavilion 3, P3-#1816

[Mark Ibrahim @marksibrahim]

Come learn about computer-use agents with OpenApps, oral at #ICLR2026 in Rio 🇧🇷on Saturday 2:35pm ET Room 204 or stop by our poster in the morning https://t.co/5gLvS3ztdb w/ @karen_ullrich

https://t.co/ohMN7s2Vo4

[Mark Ibrahim @marksibrahim]

Want to teach AI agents to use apps like humans? Get started with digital agents research using OpenApps, our new Python-based environment.

[Dr. Karen Ullrich @karen_ullrich]

I am soon heading to Rio for #ICLR2026!

It is going to be a packed week: including an oral presentation of OpenApps, our work on measuring how reliable UI agents really are when the apps they interact with change.

[Sharut Gupta @sharut_gupta]

1/n Can LLMs learn to reason on hard benchmarks like AIME and GPQA purely through context, without SFT, RL, or any weight updates?

Turns out… Yes! And it can have strong performance while being highly efficient

Paper: https://t.co/mEoaIst6cX
Blog: https://t.co/lZli7qY4Jz

[Jack Morris @jxmnop]

at long last, the final paper of my phd

🧮 Learning to Reason in 13 Parameters 🧮

we develop TinyLoRA, a new ft method. with TinyLoRA + RL, models learn well with dozens or hundreds of params

example: we use only 13 parameters to train 7B Qwen model from 76 to 91% on GSM8K 🤯

[Mark Ibrahim @marksibrahim]

@fujikanaeda Related finding showing a single character can break LLM evals: https://t.co/JQr0JDfVWP

[Mark Ibrahim @marksibrahim]

One can manipulate LLM rankings to put any model in the lead—only by modifying the single character separating demonstration examples. Learn more in our new paper https://t.co/D8CzSpPxMU

w/ Jingtong Su, Jianyu Zhang, @karen_ullrich , and Léon Bottou.
1/3 🧵

[Basile Terver @BasileTerv987]

My first PhD paper is out! 🎓

"What Drives Success in Physical Planning with Joint-Embedding Predictive World Models?"

tl:dr: JEPA-WMs for robotics: learn dynamics on top of visual encoders, optimize actions towards goal 👇

w/ @JimmyTYYang1, Jean Ponce, @AdrienBardes, @ylecun

[Dr. Karen Ullrich @karen_ullrich]

Release Day 🎉

Meet OpenApps — a pure-Python, open-source ecosystem for stress-testing UI agents at scale.

Runs on a single CPU. Generates thousands of unique UI variations. And it reveals just how fragile today’s SOTA agents are.

(Yes, even GPT-4 and Claude struggle.)

[Mark Ibrahim @marksibrahim]

@browsercompany @fasthtml @openstreetmap in collaboration with the excellent research team at FAIR: @karen_ullrich Jingtong Su @randall_balestr @_amirbar Claudia Shi, Arjun Subramonian, Nikolaos Tsilivis, Ivan Evtimov, adn @KempeLab

[Mark Ibrahim @marksibrahim]

Want to teach AI agents to use apps like humans? Get started with digital agents research using OpenApps, our new Python-based environment.

[Mark Ibrahim @marksibrahim]

built on top of excellent framework thanks to @browsercompany @fasthtml @openstreetmap

[Dr. Karen Ullrich @karen_ullrich]

Stop by the Meta booth tomorrow, Wednesday Dec 3rd at #NeurIPS in San Diego! 🤖📱

We demo our new research environment, OpenApps, for digital agents. Generate thousands of app versions to train and evaluate multimodal agents to use apps like humans do.

Not attending? Stay tuned