Luca Zappella

[Luca Zappella @luca_zapp]

What we're looking for:
• A track record of publishing at top-tier ML conferences and journals
• Hands-on experience across every stage of a research project

If this sounds interesting:
Know someone who'd be a great fit? Tag them or share this post. 🙏

[Luca Zappella @luca_zapp]

🚀 We're hiring a Research Scientist, Machine Learning Research (MLR), Barcelona ☀️

👉 Apply here: https://t.co/S0ebqBFUcx

#MachineLearning #MLResearch #AI #Hiring #Apple #Barcelona #ResearchScientist #DeepLearning

[Luca Zappella @luca_zapp]

Our MLR team is looking for someone who's excited about understanding, challenging, and re-imagining machine learning, pushing the boundaries of what we know and turning that knowledge into new capabilities across Apple's products.

[Luca Zappella @luca_zapp]

Papers capture results. What they don't capture is the immense effort this amazing team has invested, the countless brainstorms, late-night debugging sessions, and honest debates that shaped every decision. I am grateful for being part of this team, you are amazing.

[Jason Ramapuram @jramapuram]

None of this would have been possible without an extraordinary team from . Huge shoutout to everyone that pushed extrememly hard to make this happen: @LouisBAlgue, @victorturrisi, @kayembruno, @prlz77, @LokeshBoomi, Nikhil Bhendawade, @AmitisShidani1, @JorisPelemans, @theo_olausson, Devon Hjelm, Paul Dixon, @joaomonteirof, @PierreAblin, Vishnu Banna, @ArnoBlaas, Nick Henderson, Kari Noriy, @danbusbridge, @jmsusskind, Marco Cuturi, Irina Belousova, @luca_zapp, and Russ Webb.

Particular shout out to @LouisBAlgue and @victorturrisi for the many sleepless nights working together!

This was a genuine team effort - the kind of collaborative, careful science I feel fortunate to be part of.

Paper: https://t.co/ZJHmk5OLsl

[Jason Ramapuram @jramapuram]

None of this would have been possible without an extraordinary team from . Huge shoutout to everyone that pushed extrememly hard to make this happen: @LouisBAlgue, @victorturrisi, @kayembruno, @prlz77, @LokeshBoomi, Nikhil Bhendawade, @AmitisShidani1, @JorisPelemans, @theo_olausson, Devon Hjelm, Paul Dixon, @joaomonteirof, @PierreAblin, Vishnu Banna, @ArnoBlaas, Nick Henderson, Kari Noriy, @danbusbridge, @jmsusskind, Marco Cuturi, Irina Belousova, @luca_zapp, and Russ Webb.

Particular shout out to @LouisBAlgue and @victorturrisi for the many sleepless nights working together!

This was a genuine team effort - the kind of collaborative, careful science I feel fortunate to be part of.

Paper: https://t.co/ZJHmk5OLsl

[David Grangier @GrangierDavid]

#NeurIPS2025 Mixing different datasets to train your LLM?
✨ We can help you find the perfect blend!
📈 Few small-model experiments → scaling law fit → your optimal mixture.
🎯 Easy + efficient.
Chat with us 💬 Poster #3414. Thu, Dec 4, 11am

[Luca Zappella @luca_zapp]

Our new paper “𝗣𝗮𝗿𝗮𝗥𝗡𝗡: 𝗨𝗻𝗹𝗼𝗰𝗸𝗶𝗻𝗴 𝗣𝗮𝗿𝗮𝗹𝗹𝗲𝗹 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗼𝗳 𝗡𝗼𝗻𝗹𝗶𝗻𝗲𝗮𝗿 𝗥𝗡𝗡𝘀 𝗳𝗼𝗿 𝗟𝗟𝗠𝘀” is out
📄 Paper: https://t.co/QSj2pzs0ML
💻 Code: https://t.co/J9PH1ZgaGh
WIth @FedericoDa40495 @prlz77 @XaviSuauC Miguel Sarabia

[Federico Danieli @FedericoDa40495]

𝗣𝗮𝗿𝗮𝗥𝗡𝗡: 𝗨𝗻𝗹𝗼𝗰𝗸𝗶𝗻𝗴 𝗣𝗮𝗿𝗮𝗹𝗹𝗲𝗹 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗼𝗳 𝗡𝗼𝗻𝗹𝗶𝗻𝗲𝗮𝗿 𝗥𝗡𝗡𝘀 𝗳𝗼𝗿 𝗟𝗟𝗠𝘀
For years, we’ve given RNNs for doomed, and looked at Transformer as 𝘁𝗵𝗲 LLM—but we just needed better math
📄https://t.co/lFQrUEfEvZ
💻https://t.co/Lg7gbcwgFU

[Luca Zappella @luca_zapp]

Check out our new work on conditioning pre-trained generative models via activation steering. LineAS has been accepted at NeurIPS 2025. Code and paper are online:

💻 https://t.co/MJSZHe9lq0
📄 https://t.co/5owafAn77X

[Pau Rodríguez @prlz77]

🚀 Excited to share LinEAS, our new activation steering method accepted at NeurIPS 2025! It approximates optimal transport maps e2e to precisely guide 🧭 activations achieving finer control 🎚️ with ✨ less than 32 ✨ prompts!

💻https://t.co/IdZOpwtFXC
📄https://t.co/sfPHk5sT2B

[Luca Zappella @luca_zapp]

Our latest work on understanding Mamba models will be presented at #icml2025! Great work by last summer intern @TeresaNHuang together with Miguel Sarabia,
@amoudgl, @prlz77, and Federico Danieli.

[Teresa Huang @TeresaNHuang]

Is the mystery behind the performance of Mamba🐍  keeping you awake at night? We got you covered! Our ICML2025 paper demystifies input selectivity in Mamba from the lens of approximation power, long-term memory, and associative recall capacity.
https://t.co/dWDYyIWLzt

[Pavankumar Vasu @PavankumarVasu]

Excited to share code & models for FastVLM — our blazing-fast Vision-Language Model appearing at #CVPR2025

Run it on-device with inference code optimized for Apple Silicon using #mlx.
Code: https://t.co/zrYytwr9N1

Updated paper & results coming soon. Stay tuned! 👀

[Luca Zappella @luca_zapp]

New blog post that explains our work on Controlling Diffusion and LLMs using steering and optimal transport:
https://t.co/RTZyeZkIJc

This work will be presented at ICLR2025 in Singapore. See you there!

[Pau Rodríguez @prlz77]

Our work on fine-grained control of LLMs and diffusion models via Activation Transport will be presented @iclr_conf as spotlight✨Check out our new blog post https://t.co/dAJQtcETNX

[Rin Metcalf Susa @RinMetcalfSusa]

🚀 We're hiring an ML Researcher! 🚀

If you're an expert in LLM alignment & personalization and want to work on a world-class research team, apply here 👉 https://t.co/dODHx2IHeF

Know someone who’d be a great fit? Tag them! #MachineLearning #AI #Apple

[Arwen Bradley @ArwenBradley]

When does composition of diffusion models “work”? Prior work (Du et al., 2023; Liu et al., 2022) has shown that composition via linear score combination can sometimes compose concepts like “dog” and “oil painting”, but why? Does it always work? https://t.co/667SNBXKVa

[Samira Abnar @samira_abnar]

🚨 One question that has always intrigued me is the role of different ways to increase a model's capacity: parameters, parallelizable compute, or sequential compute?

We explored this through the lens of MoEs:

[MINT Workshop @ NeurIPS2024 @MINT_workshop]

Two days until our @NeurIPS workshop on Model Interventions on Foundation Models on Sunday!!

Our second talk features @viegasf, Professor @Harvard and co-lead of @Google's PAIR initiative presenting on user-centric interpretability.

https://t.co/eD0z0pcVxS

[Pau Rodríguez @prlz77]

Interested in how interpretability can be used for alignment? Join us at the workshop on foundation model interventions the 15th of December at @NeurIPSConf!

[Luca Zappella @luca_zapp]

Our new work on adding fine-grained control to LLMs and Diffusion Models via optimal transport (no retraining of the model needed!) 🔥

[Pau Rodríguez @prlz77]

Thrilled to share the latest work from our team at @Apple where we achieve interpretable and fine-grained control of LLMs and Diffusion models via Activation Transport 🔥

📄 https://t.co/TYlwxarrWx
🛠️ https://t.co/gciUcwRNqd

1/9 🧵

[Luca Zappella @luca_zapp]

3 Internship positions in @Apple ML Research in Barcelona!

[Pau Rodríguez @prlz77]

I’m thrilled to announce 3 #internship openings @Apple ML Research in beautiful ☀️ #Barcelona ☀️ for 2025! Two internships on Generative Models (GM), Controllability, Interpretability, and Model Editing; and one on GM &🔈Spatial Audio. Apply: https://t.co/RG6OobIvL3

Details 🧵

[Pau Rodríguez @prlz77]

I’m thrilled to announce 3 #internship openings @Apple ML Research in beautiful ☀️ #Barcelona ☀️ for 2025! Two internships on Generative Models (GM), Controllability, Interpretability, and Model Editing; and one on GM &🔈Spatial Audio. Apply: https://t.co/RG6OobIvL3

Details 🧵

[Jason Ramapuram @jramapuram]

Enjoy attention? Want to make it ~18% faster? Try out Sigmoid Attention. We replace the traditional softmax in attention with a sigmoid and a constant (not learned) scalar bias based on the sequence length.

Paper: https://t.co/49FeKwLAE0
Code: https://t.co/b0cp49pXKX

This was an amazing collaboration with some great researchers at Apple: Federico Danieli, @EeshanDhekane, @FlorisWeers, @danbusbridge, @PierreAblin, Tatiana Likhomanenko, Jagrit Digani, Zijin Gu, @AmitisShidani1 and Russ Webb.

More details in 🧵below.