Apple will be presenting new research from July 13th to 19th at the International Conference on Machine Learning (ICML 2025) held in Vancouver, Canada. We are proud to once again sponsor cutting-edge research such as machine vision, computational biology, robotics and more, with a focus on presenting and publishing cutting-edge research in machine learning, artificial intelligence, statistics, data science and related applications. Below is an overview of Apple's participation in ICML 2025.
Jump to section:
schedule
Please stop by the Apple booth (#307) at the Vancouver Convention Center during exhibition hours. All times are listed in PT (Vancouver time):
- Monday, July 14th: 4pm – 8pm
- Tuesday, July 15th: 9:30am to 6:00pm
- Wednesday, July 16th: 9:30am to 6:00pm
- Thursday, July 17th: 9:30am to 1pm
Sunday, July 13th
Monday, July 14th
- Affinity Group
- Latinx in AI Workshop
- 3pm – 4pm
West Conference Room 118-120 - Lauren Araujo and Miguel Saravia del Castillo will represent Apple in the LXAI Mentor Panel Discussion.
Tuesday, July 15th
- Poster
- “COMMVQ: Commuting Vector Quantization for KV Cache Compression”
- 4:30pm to 7pm
Poster Session 2 East, #e-2711, East Exhibition Hall AB - Junyan Li (University of Massachusetts Amherst), Yang Chan (Massachusetts Institute of Technology), Muhammad Yusuf Hassan (University of Massachusetts Amherst), Tarha Chafekar (University of Massachusetts Amherst), Tian Kai (Princeton University), Chong Wang, Chuan Gang (University of Massachusetts Amherst)
- Poster
- “flextok: Image resampling into flexible length 1D token sequences”
- 4:30pm to 7pm
Poster Session 2 East, #e-3010, East Exhibition Hall AB - Roman Bachmann, Jesse Araldac, David Mizrahi, Enrico Fini, Ozhan Fati Kerr (EPFL), Elmira Amilou, Araeldin El Nubie, Amir Zamir (EPFL), Afsin Dehagan
Wednesday, July 16th
- Poster
- “Distillation method”
- 11am – 1:30pm
Poster Session 3 East, #E-2310, East Exhibition Hall AB - Dan Basbridge, Jason Ramapuram, Floris Weirs, Fried Fire (Oxford University), Russ Webb, Etai Litwin
- Poster
- “Pruning following the instruction of large-scale language models”
- 11am – 1:30pm
Poster Session 3 East, #E-2711, East Exhibition Hall AB - Bairu Hou (UCSB), Qibin Chen, Jianyu Wang, Guoli Yin, Chong Wang, Nan Du, Ruoming Pang, Shiyu Chang (UCSB), Tao Lei
- Poster
- “Are your model pretty certain? A detailed approach to LLM fairness.”
- 11am – 1:30pm
Poster Session 3 East, #E-905, East Exhibition Hall AB - Oliver Wang (Carnegie Mellon University), NIV Shibakumar, Farah Arif Kahn (New York University), Lynn Metcalf Susa, Adam Golinsky, Natalie McCraz, Barry Theobaldo, Luca Zappera, Nick Apostrov
- Poster
- “Proxy-FDA: Don't forget to use proxy-based feature delivery alignment for the Vision Foundation model.”
- 11am – 1:30pm
Poster Session 3 East, #E-1309, East Exhibition Hall AB - Chen Fang, Skylar Set, Hadi Poor Ansari, Meldad Farajtabal, Raviteja Vemrapali, Faltash Fagri, Onsel Tsel, Barry Theobaldo, Josh Suskindin
- Affinity Group
- Machine Learning Women (WIML)
- 1pm – 2pm
West Conference Room 211-214 - Lucy Perrotta, Tatiana Rikhomanenko and Wasifa Jamal represent Apple in the WIML Mentorship Round Table Luncheon
- Poster
- “Learn to route LLMS with confidence tokens”
- 4:30pm to 7pm
Poster Session 4 East, #e-1208, East Exhibition Hall AB - Allen Chuan (Rice University), Plassha Kemswara Salma, Palikshit Gopalan, John Boscio, Sarah Boluki, Xia (Ben) Hu (Rice University), Helen Zhou
Thursday, July 17th
- Poster
- “Quantspec: Self-Specific Decoding Using a Layered Quantized KV Cache”
- 11am – 1:30pm
Poster Session 5 East, #E-2608, East Exhibition Hall AB - Rishabh Tiwari (UC Berkeley), Haocheng Xi (UC Berkeley), Aditya Tomar (UC Berkeley), Coleman Hooper (UC Berkeley), Sehoon Kim (UC Berkeley), Max Horton, Mahyar Najibi, Michay W. Mahoney (UC Berkeley), Keutzer Amir Gholaminejad (UC Berkeley)
- Poster
- “Understanding input selectivity in mamba: Impact on approximation, memorization, and associative recall ability”
- 11am – 1:30pm
Poster Session 5 East, #e-2312, East Exhibition Hall AB - Teresa Juan (Johns Hopkins University), Abhinav Muzil (Concordia University, Montreal, Canada), Pau Rodriguez Lopez, Miguel Saravia del Castillo, Luca Zappera, Federico Danieli
- Oral
- “Normalization flow is a capable generative model.”
- 4pm – 4:15pm
West Ballroom A, Oral 6B Deep Learning Architecture - Shuangfei Zhai, Ruixiang Zhang, Preetum Nakkiran, David Berthelot, Jiatau Gu, Huangjie Zheng, Tianrong Chen, Miguel Angel Bautista Martin, Navdeep Jaitly, Josh Susskind
- Poster
- “Pre-training localized language images for contrast”
- 4:30pm to 7pm
Poster Session 6 East, #e-3209, East Exhibition Hall AB - Hong-You Chen, Jeff Lai, Haotian Zhang, Angie Wang, Marcin Eichner, Keen You, Meng Cao, Bowen Zhang, Yinfei Yang, Zhe Gan
- Poster
- “Normalization flow is a capable generative model.”
- 4:30pm to 7pm
Poster Session 6 East, #e-2911, East Exhibition Hall AB - Shuangfei Zhai, Ruixiang Zhang, Preetum Nakkiran, David Berthelot, Jiatau Gu, Huangjie Zheng, Tianrong Chen, Miguel Angel Bautista Martin, Navdeep Jaitly, Josh Susskind
- Poster
- “Sturdy autonomy emerges from self-play.”
- 4:30pm to 7pm
Poster Session 6 West, #W-821, West Exhibition Hall B2-B3 - Marco Kusmano Towner, David Hafner, Alex Hertzberg, Brody Houval, Alexei Petrenko, Eugene Vinnitsky, Eric Wijmans, Taylor Killian, Stuart Bauers, Ozan Sen, Philip Clairehenbuell, Vladren Koltunn
Friday, July 18th
Saturday, July 19th
- Workshop paper
- “Lightweight safety guardrails via synthetic data and hostile training with RL guides”
- Generation Model Workshop Data
West Ballroom A, 8:30am to 5pm - Rachel Li, Riyaaz Shaik, Noyan Tokgozoglu, Aleksei ilin, Amber MA, Gor Matevosyan, Suhaa Dada, James Rae
- Workshop paper
- “Lightweight safety guardrails via synthetic data and hostile training with RL guides”
- Workshop on Reliable Responsible Basic Models 2025
8:50am to 6pm, West Ballroom c - Rachel Li, Riyaaz Shaik, Noyan Tokgozoglu, Aleksei ilin, Amber MA, Gor Matevosyan, Suhaa Dada, James Rae
Booth programming and demo
Please check out Apple's Booth (#307) at Vancouver Convention Centre during exhibition hours.
Technical demo
- demo
- MLX
- MLX is a flexible array framework optimized for Apple silicon and brought to you by Apple Machine Learning Research. This allows training and inference of any complex model on Apple silicon-powered devices. We will showcase image generation using fine tuning 7B parameter LLM on iPhone, large spreading models on iPad, and text generation using numerous major language models in M2 Ultra Mac Studio.
- Monday, July 14th: 4pm – 8pm
- Tuesday, July 15th: 9:30am to 12:00pm, 2:00pm – 4:00pm
- Wednesday, July 16th: 9:30am to 12:00pm, 2:00pm – 4:00pm
Accepted papers
- “ASWC: Adaptive Scalable Weight Compression for Large-Scale Language Models”
- Alvin Wang, Colorado Reed, Jason Wang, Wenki Chan, Dowai Wang, Binazil Karimzadeh, Sam XU
- “Audiovisual language model for controllable talking head synthesis”
- Karren Yang, Sai Srujana Buddi, Chun-Liang Li, Richard Bai, Vikram Mitra, Zak Aldeneh, Tatiana Likhomanenko, Akash Mehra, Navdeep Jaitly, Oncel Tuzel
- “COMMVQ: Commuting Vector Quantization for KV Cache Compression”
- Junyan Li (University of Massachusetts Amherst), Yang Chan (Massachusetts Institute of Technology), Muhammad Yusuf Hassan (University of Massachusetts Amherst), Tarha Chafekar (University of Massachusetts Amherst), Tian Kai (Princeton University), Chong Wang, Chuan Gang (University of Massachusetts Amherst)
- “Distillation method”
- Dan Basbridge, Jason Ramapuram, Floris Weirs, Fried Fire (Oxford University), Russ Webb, Etai Litwin
- “Online Lag: Enhance Long Context Understanding in LLMS using Searched Mechanisms”
- Shishir Patil (University of California, Berkeley), Alvin Wang, Colorado Reed, Joseph E. Gonzalez (University of California, Berkeley), Pranaf Ramesh (University of California, Berkeley), Aeon Stoika (University of California, Berkeley), QI Shan
- “Sturdy autonomy emerges from self-play.”
- Marco Kusmano Towner, David Hafner, Alex Hertzberg, Brody Houval, Alexei Petrenko, Eugene Vinnitsky, Eric Wijmans, Taylor Killian, Stuart Bauers, Ozan Sen, Philip Clairehenbuell, Vladren Koltunn
Accepted workshop papers
- “ksparsecache: KV cache compression via K-SPARSE autoencoder for ks-decomposed LLM serving”
- Pulkit Agrawal, Minsik Cho, Arnav Kundu
- ES-FOMO III: 3rd Workshop on Efficient Systems for Basic Models 2025
- “Lightweight safety guardrails via synthetic data and hostile training with RL guides”
- Rachel Li, Riyaaz Shaik, Noyan Tokgozoglu, Aleksei ilin, Amber MA, Gor Matevosyan, Suhaa Dada, James Rae
- Multiple workshops (Data from Generation Model 2025, DataWorld: Unified Data Curation Framework across Domains 2025, Workshop on Reliable Responsible Basic Models 2025)
Acknowledgments
Natalie Schluter is an ICML workshop chair.
Kunal Talwar, Marco Cuturi, Ronan Collobert, Samy Bengio and Vladlen Koltun are senior area chairs at ICML.
Byeongjoo Ahn, Chun-Liang Li, Louis Bethune, Luca Zappella, Miguel Angel Bautista Martin, Pau Rodriguez Lopez, Preetum Nakkiran, Shuangfei Zhai, Tatiana Likhomanenko and Xavier Suau Cuadros are area directors of ICML.
Dan Busbridge, Michael Kirchhof, Qingqing Cao, and Shu W. are recognized as top reviewers of ICML.
Aakash Goel, Antoine Wehenkel, Arno Blaas, Arwen Bradley, Barry Theobald, Dan Busbridge, Devon Hjelm, Eugene Ndiaye, Fahad Kamran, Fartash Faghri, Hadi Pour Ansari, Hillal Asi, Huangjie Zheng, Katherine Tshen Miguel Saravia del Castillo, Niv Shivakumal, Pavan Kumar Anasosar Vas, Philip Klehehenbuhl, Philip Chilensky, Pierre Ablin, Qingqing cao, Lynn Metaluff Susa, Tianlongchen, Vimal Tilak, Vitai Feldman, Xiaom, Zhang and Zhe Gan are meeting reviewers for ICML.
Kriti Goyal is a reviewer for Latinx, an AI workshop.
