←
Mark Ibrahim92 posts
Mark Ibrahim avatar

Mark Ibrahim

@marksibrahim

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

◷ Joined December 2012▧ 9 Photos and videos
Rohan Paul avatar
⟳ Mark Ibrahim retweeted
Rohan Paul@rohanpaul_ai·Oct 29

Nice new paper on continuous improvement of AI agents from Princeton and Meta Finds that agents repeat fewer mistakes when each fix is saved as a LoRA adapter and reused on similar tasks. Agents can fix a mistake within a session, then often repeat it in the next one. Notes in context get forgotten, and full fine-tuning is costly and can disrupt other skills. With Self-Direction, a failing agent reflects on its run and trains the lesson into a LoRA adapter about 0.02% of the model's size. A router later loads the closest adapter when the agent fails a similar task. With Qwen3.6-27B, task success on GAIA rose from 38.8% to 55.8%, beating in-context reflections and a Skills.md file. Across 185 earlier mistakes, 67.3% of repairs still worked by the third encounter, against 42.9% for plain retrying. If your agents keep making the same mistakes, save each fix as a small adapter they can reload, not only as prompt notes.

Post attachment
Mark Ibrahim avatar
Mark Ibrahim@marksibrahim·Oct 26

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.

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 💪🏽 🌐 self-direction.github.io/ 🧵1/n

0:24
Yuval Kansal avatar
⟳ Mark Ibrahim retweeted
Yuval Kansal@yuvalkansal·Oct 26

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 💪🏽 🌐 self-direction.github.io/ 🧵1/n

0:24
Mark Ibrahim avatar
Mark Ibrahim@marksibrahim·Oct 15

Muse Spark Open, an open-weight frontier model that can run on two of the latest Mac Studios: x.com/AIatMeta/status/211077… Proud of our team’s contributions! @candacerossio @eshikasax @adinamwilliams + many others

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

Post attachmentPost attachment
Reyhane Askari avatar
⟳ Mark Ibrahim retweeted
Reyhane Askari@ReyhaneAskari·Apr 24

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

New @AIatMeta paper explains when a smaller, curated dataset beats using everything. Standard training wastes effort because many examples are redundant or wrong. They formalize a label generator, a pruning oracle, and a learner. From this, they derive exact error laws and sharp regime switches. With a strong generator and plenty of data, keeping hard examples works best. With a weak generator or small data, keeping easy examples or keeping more helps. They analyze 2 modes, label agnostic by features and label aware that first filters wrong labels. ImageNet and LLM math results match the theory, and pruning also prevents collapse in self training. ---- Paper – arxiv. org/abs/2511.03492 Paper Title: "Why Less is More (Sometimes): A Theory of Data Curation"

Post attachment
Mark Ibrahim avatar
Mark Ibrahim@marksibrahim·Apr 21

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 iclr.cc/virtual/2026/oral/10… w/ @karen_ullrich x.com/marksibrahim/status/19…

Post attachment

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

2:10
Dr. Karen Ullrich avatar
⟳ Mark Ibrahim retweeted
Dr. Karen Ullrich@karen_ullrich·Apr 15

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.

Jack Morris avatar
⟳ Mark Ibrahim retweeted
Jack Morris@jxmnop·Feb 5

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 🤯

Post attachment
Basile Terver avatar
⟳ Mark Ibrahim retweeted
Basile Terver@BasileTerv987·Jan 12

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

1:08
Dr. Karen Ullrich avatar
⟳ Mark Ibrahim retweeted
Dr. Karen Ullrich@karen_ullrich·10 Dec 2025

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.)

2:10