David Fan

I thought the path to variable-length video was frame-wise autoregressive with complex forcing schedules, but I was wrong! The solution is simple! Flowception, using frame insertions to model any-order video generation.✅ Arxiv: arxiv.org/abs/2512.11438 Project page: flowception-meta.github.io Great collaboration with Jakob, @inthebrownbag, and @RickyTQChen on this work, and led by Tariq!

In #ContinualLearning, we have long assumed extreme memory constraints. But in today's era of large models where compute is the real bottleneck, is catastrophic forgetting still our biggest problem? Our #ICLR2026 paper, "Forget Forgetting: Continual Learning in a World of Abundant Memory", challenges this view. We found that in practical settings with abundant memory, models actually struggle more with a loss of plasticity rather than forgetting. They become biased toward past data and fail to absorb new knowledge. To address this, we introduce Weight Space Consolidation. This lightweight method uses rank-based parameter initialization (using merging) to restore plasticity and weight averaging to maintain stability. The result is high efficiency and performance in both #LLM continual instruction tuning and vision models, avoiding the massive cost of full retraining. While I cannot attend the conference in Rio de Janeiro in person, the first author will be presenting our poster. If you are working on the practical limits of Continual Learning or LLMs, please stop by our session for a discussion. Presentation Details: ICLR 2026, Poster Session 5, Pavilion 3 (Saturday, April 25 at 09:30 EDT) Project Page: sites.google.com/view/forget-fo… GitHub: github.com/umamicode/weig…

Congrats Dr. Tong! Really glad to be a part of your PhD journey @TongPetersb

Introducing DexWM: Dexterous Manipulation World Model tl;dr: train on human videos; fine-grained actions; hand-consistency loss; DexWM+MPC -> zero-shot dexterous manipulation w/ @_amirbar, @DavidJFan, @JimmyTYYang1, @GaoyueZhou, P. Krishnamurthy, M. Rabbat, F. Khorrami, @ylecun

The code for DexWM is now publicly available: github.com/facebookresear…. The repository includes the full training and evaluation pipelines, along with custom dexterous manipulation datasets generated in RoboCasa, making it easy to reproduce our results and build on top of this work.

Advanced Machine Intelligence (AMI) is building a new breed of AI systems that understand the world, have persistent memory, can reason and plan, and are controllable and safe. We’ve raised a $1.03B (~€890M) round from global investors who believe in our vision of universally intelligent systems centered on world models. This round is co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, along with other investors and angels across the world. We are a growing team of researchers and builders, operating in Paris, New York, Montreal and Singapore from day one. Read more: amilabs.xyz AMI - Real world. Real intelligence.