Twitxy Saved

Fartash Faghri

Verified account This account is verified. Learn more

@FartashFg

ML Research @Apple. @UofT PhD.
Toronto,Canada fartashf.github.io Joined November 2013
71 Following 1.2K Followers 227 Posts

Latest posts 5

Fartash Faghri Verified account @FartashFg
📣 We’re hiring PhD research interns at Apple ML Research! 
Focus: Efficient #multimodal models, video & memory 📹
Goal: Publish high-impact research 📝
Duration: 4–10 months (Nov '26 – Sept '27)

Building on our recent work in fast vision-language models: machinelearning.apple.com/research/fast-…
Apply at the link below and be sure to include "#efficient-ml" in your resume/cover letter so we can spot your application! 👇
jobs.apple.com/en-us/details/…
machinelearning.apple.com FastVLM: Efficient Vision Encoding for Vision Language Models Vision Language Models (VLMs) enable visual understanding alongside textual inputs. They are typically built by passing visual tokens from a…
6 replies 27 reposts 252 likes 23K views View on X
Fartash Faghri Verified account @FartashFg
In the era of continued pretraining and continued fine-tuning, loss of plasticity means leaving future gains on the table. We need a better theoretical understanding of loss of plasticity. See a great thread unpacking the dynamics. 👇
#ICLR2026 #ContinualLearning #DeepLearning
Amir Joudaki Verified account @AmirJoudaki Apr 25
Neural nets don’t just forget. Sometimes, after long training, they lose the ability to learn at all.

In our #ICLR2026 poster, we model Loss of Plasticity as gradient dynamics trapped in invariant manifolds: 🔴 frozen units, 🔵 cloned units.

The video makes the traps visible.
View quoted post on X
0 replies 4 reposts 12 likes 2.4K views View on X
Fartash Faghri Verified account @FartashFg
Attending #ICLR2026! Feel free to message me to chat about efficient multimodal models, or come find me at:
🗣️ Chairing Oral Session 4C: Vision Language Models III
Fri 24 Apr | 3:15 PM - 4:45 PM
📊 MobileCLIP2 (DFNDR 2B/12M released, links below👇)
Sat 25 Apr | 10:30 AM - 1:00 PM | Poster Session 5, Pavilion 4, #3713
🪧 Apple Booth, Sat 25 Apr | 1:30 PM - 3:30 PM
📊Barriers for Learning in an Evolving World: Mathematical Understanding of Loss of Plasticity
Sat 25 Apr | 3:15 PM - 5:45 PM | Poster Session 6, Pavilion 4, #4202
📊 Data-Centric Lessons To Improve Speech-Language Pretraining
Sat 25 Apr | 3:15 PM - 5:45 PM | Poster Session 6, Pavilion 3, #1418

Come work with our MLR team on efficient ML (message me if interested!): jobs.apple.com/en-us/details/…
Apple @ ICLR 2026: machinelearning.apple.com/updates/apple-…
DFNDR-2B: huggingface.co/datasets/apple…
DFNDR-12M: huggingface.co/datasets/apple… huggingface.co/datasets/apple…
Fartash Faghri Verified account @FartashFg
Join us tomorrow for CCFM workshop, Upper Level Room 25ABC, Sunday, December 7th 2025, 8am-5pm.
neurips.cc/virtual/2025/l…
Fartash Faghri Verified account @FartashFg Jul 7, 2025
Is your AI keeping Up with the world?

Announcing #NeurIPS2025 CCFM Workshop: Continual and Compatible Foundation Model Updates
When/Where: Dec. 6-7 San Diego
Submission deadline: Aug. 22, 2025. (opening soon!)
sites.google.com/view/ccfm-neur…

#FoundationModels #ContinualLearning
View quoted post on X