AI researcher @Apple (opinions are my own)
Cupertino
Joined August 2011
- Tweets112
- Following130
- Followers1.9K
- Likes232
Oncel Tuzel retweeted
Introducing a routing control framework to train agentic MoEs: aligning turn-level experts with agent actions while encouraging token-level consistency leads to major performance gains.
arxiv.org/pdf/2610.07332
@lblaoke Ting-Yao Hu @cydhsieh @schowdhury671 @OncelTuzel
Oncel Tuzel retweeted
So proud of the MLX team!
Muse Spark Open, on your desk: 816B params at 36+ tokens/sec on 4 Mac Studio computers.
We've been hard at work on better M5 GPU Neural Accelerator support, a smoother Thunderbolt experience (including multi-cable meshes), and kernel optimizations for MoE architectures.
Oh — and MLX LM (mlx_lm.generate, mlx_lm.server) now supports the vision modality for Muse Spark Open!
More on using MLX for distributed inference of frontier-scale models on Mac Studio:
machinelearning.apple.com/re…
Come and work with us!
📣 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/re…
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…
Oncel Tuzel retweeted
We let an agent self-improve by exploring and internalizing ”on this sort of task, keep this sort of thing in mind“ bits of self-feedback. It learned to solve tasks where the original policy failed 128 attempts in a row and GRPO flatlined.
Paper: arxiv.org/abs/2609.37633
🧵1/8
Oncel Tuzel retweeted
We’re excited to share that @nuance_ai has raised a $50M Series A led by @lightspeedvp, joined by @Accel, @spc, @nvidia, and @definevc!
Much of human communication happens beyond words. Before we can speak, we learn to read faces, return smiles, follow gestures, and sense when someone is listening. That visual and emotional vocabulary stays with us for life.
At Nuance Labs, we’re building a foundation model that understands and responds to the full spectrum of human emotion. A single full-duplex audio-visual model that sees, listens, and responds in real time, with the natural give-and-take of a conversation with a friend or colleague.
We believe the best interface with a machine is the one we’ve practiced since birth.
Join us!
Oncel Tuzel retweeted
Tired of building complex environments to train AI agents? 🛑
Check out our environment-free way to generate high-quality agentic trajectories
#DigitalWorldModeling
arxiv.org/abs/2607.16900
@seanie_12 @schowdhury671 @chaojiang06 @cydhsieh Ting-Yao Hu @alexttoshev @OncelTuzel
Our team at Apple MLR is presenting 4 papers at #CVPR2026! If you're attending, stop by and chat with the authors about our research. 🧵
📌 Velox — Malik et al.
apple.github.io/ml-velox/
Jun 6, 11:45 AM MDT
📌 AMusE — Chowdhury et al.
arxiv.org/abs/2512.16250
Jun 6, 4:45 PM MDT
📌 TrajTok — Zheng et al.
arxiv.org/abs/2602.22779
Jun 7, 11:45 AM MDT
📌 VSAS-Bench (CVPR Findings) — Vasu et al.
arxiv.org/abs/2604.07634
Jun 7, 7:30 AM MDT
We're also hiring researchers in Efficient ML at Apple MLR.
jobs.apple.com/en-us/details…
#Apple #EfficientML #GenAI #MultimodalAI
I’ll be speaking at the EDLCV and EDGE workshops at CVPR 2026 on June 3 in Colorado.
Join us if you're interested in efficient ML, on-device AI, and recent research from Apple.
EDLCV: ecv-workshop.github.io/
EDGE: cvpr26-edge.github.io/
Great work on 4D representation learning from Anagh’s internship at @Apple, with exciting applications in video-to-4D, 3D tracking, and simulation.
📢📢📢 Velox 🚀: Learning Representations of 4D Geometry and Appearance
In our #CVPR2026 paper, we introduce a method for learning a native 4D representation, useful for many downstream tasks, such as video-to-4D, 3D tracking, cloth simulation, and others!
🌐: apple.github.io/ml-velox
📝: arxiv.org/abs/2605.04527
Excited to share that we’ve open-sourced LiTo (ICLR 2026) code + models from @Apple!
Interactive image-to-3D generation from images.
• Apple Silicon demo via MLX
• Full training code
GitHub: github.com/apple/ml-lito
Paper: arxiv.org/abs/2603.11047
#Apple #MLX #3D #AI
Our research group at @Apple is presenting 6 papers at ICLR 2026. If you're attending, come meet our colleagues @FartashFg, @raviteja_vemu, @ChangRick, Cem Koc, Ting-Yao Hu.
We’re also looking for strong researchers for an Efficient ML Research
jobs.apple.com/en-us/details…
Thread 👇
Learning to Reason for Hallucination Span Detection
Hsuan Su et al.
iclr.cc/virtual/2026/poster/…
Barriers for Learning in an Evolving World: Mathematical Understanding of Loss of Plasticity
Amir Joudaki et al.
iclr.cc/virtual/2026/poster/…
Data-Centric Lessons to Improve Speech-Language Pretraining
Vishaal Udandarao et al.
iclr.cc/virtual/2026/poster/…
MobileCLIP2: Improving Multi-Modal Reinforced Training (TMLR)
Fartash Faghri et al.
iclr.cc/virtual/2026/poster/…
LiTo: Surface Light Field Tokenization (ICLR 2026) — new work from Apple MLR.
LiTo learns a unified 3D representation of geometry + view-dependent appearance, capturing effects like specular highlights & Fresnel reflections, enabling high-fidelity 3D generation from single image.
Project page: apple.github.io/ml-lito/
Paper: arxiv.org/abs/2603.11047
with Jen-Hao Rick Chang, @xmzhao_, and Dorian Chan
#Apple #GenerativeAI #Research #ICLR2026
Oncel Tuzel retweeted
Autoregressive models dominate, but what if we treat multimodal generation as discrete order agnostic iterative refinement? Excited to share our systematic study on the design space of Tri-Modal Masked Diffusion Models (MDMs). We pre-trained the first Tri-Modal MDM from scratch on (text,), (image, text), and (audio, text). The same model can do ASR, TTS, T2I, captioning and native text generation.
What I'm the most proud of in this work is the scientific rigor. Over 3,500 training runs. Principled hyperparameter transfer. Honest results. Carefully controlled ablations across multiple different axis of entanglement.
A thread on our empirical findings (arXiV: arxiv.org/abs/2602.21472)
