content type eventpublished April 16, 2025
Apple is presenting new research at the annual conference on International Conference on Learning Representations (ICLR), which takes place in person in Singapore from April 24 to 28. We are proud to again sponsor the conference, which brings together the scientific and industrial research communities in deep learning. Below is an overview of Apple's participation at ICLR 2025.
Schedule
Stop by the Apple booth (#C03) in the Singapore EXPO during exhibition hours:
- Thursday, April 24: 09:30 - 17:30
- Friday, April 25: 09:30 - 17:30
- Saturday, April 26: 09:30 - 17:30
All times listed in GMT +8 (Singapore time).
Thursday, April 24
- SPOTLIGHT POSTER
- Adaptive Batch Size for Privately Finding Second-order Stationary Points
- 10:00 - 12:30, #611, Poster Session 1, Hall 3 + Hall 2B
- Daogao Liu (University of Washington), Kunal Talwar
- SPOTLIGHT POSTER
- Controlling Language and Diffusion Models by Transporting Activations
- 10:00 - 12:30, #470, Poster Session 1, Hall 3 + Hall 2B
- Pau Rodríguez, Arno Blaas, Michal Klein, Luca Zappella, Nicholas Apostoloff, Marco Cuturi, Xavier Suau
- POSTER
- Do LLMs Estimate Uncertainty Well in Instruction-Following?
- 10:00 - 12:30, #505, Poster Session 1, Hall 3 + Hall 2B
- Juyeon Heo (University of Cambridge), Miao Xiong (National University of Singapore), Christina Heinze-Deml, Jaya Narain
- POSTER
- How to Verify Any (Reasonable) Distribution Property: Computationally Sound Argument Systems for Distributions
- 10:00 - 12:30, #504, Poster Session 1, Hall 3 + Hall 2B
- Tal Herman (Weizmann Institute of Science), Guy Rothblum
- POSTER
- On the Modeling Capabilities of Large Language Models for Sequential Decision Making
- 10:00 - 12:30, #635, Poster Session 1, Hall 3 + Hall 2B
- Martin Klissarov (McGill University / Mila), Devon Hjelm, Alexander Toshev, Bogdan Mazoure
- POSTER
- Step-by-Step Reasoning for Math Problems via Twisted Sequential Monte Carlo
- 10:00 - 12:30, #276, Poster Session 1, Hall 3 + Hall 2B
- Shengyu Feng (CMU), Xiang Kong, Shuang Ma, Aonan Zhang, Dong Yin, Chong Wang, Ruoming Pang, Yiming Yang (CMU)
- SPOTLIGHT POSTER
- No Need to Talk: Training Mixture of Language Models Independently
- 15:00 - 17:30, #304, Poster Session 2, Hall 3 + 2B
- Anastasiia Filippova (EPFL), Ronan Collobert, Angelos Katharopoulos, David Grangier
- SPOTLIGHT POSTER
- Provable Uncertainty Decomposition via Higher-Order Calibration
- 15:00 - 17:30, #501, Poster Session 2, Hall 3 + 2B
- Gustaf Ahdritz (Harvard University), Aravind Gollakota, Parikshit Gopalan, Charlotte Peale (Stanford University), Udi Wieder
- POSTER
- SeedLM: Compressing LLM Weights through Seeds of a Pseudo-Random Generator
- 15:00 - 17:30, #310, Poster Session 2, Hall 3 + 2B
- Rasoul Shafipour, David Harrison, Max Horton, Jeff Marker, Houman Bedayat, Sachin Mehta (Meta), Mohammad Rastegari (Meta), Mahyar Najibi, Saman Naderiparizi
Friday, April 25
- POSTER
- Cut Your Losses in Large-Vocabulary Language Models
- 10:00 - 12:30, #217, Poster Session 3, Hall 3 + Hall 2B
- Erik Wijmans, Brody Huval, Alexander Hertzberg, Vladlen Koltun, Philipp Krähenbühl
- POSTER
- EC-DIT: Scaling Diffusion Transformers with Expert Choice Routing
- 10:00 - 12:30, #153, Poster Session 3, Hall 3 + Hall 2B
- Haotian Sun (Georgia Institute of Technology), Tao Lei, Bowen Zhang, Yanghao Li, Haoshuo Huang, Ruoming Pang, Bo Dai (Georgia Institute of Technology), Nan Du
- POSTER
- Ferret-UI 2: Mastering Universal User Interface Understanding Across Platforms
- 10:00 - 12:30, #191, Poster Session 3, Hall 3 + Hall 2B
- Zhangheng Li, Keen You, Haotian Zhang, Di Feng, Harsh Agrawal, Xiujun Li, Mohana Prasad Sathya Moorthy, Jeff Nichols, Yinfei Yang, Zhe Gan
- POSTER
- Language Models Know More Than They Show: Exploring Hallucinations From the Model's Viewpoint
- 10:00 - 12:30, #233, Poster Session 3, Hall 3 + Hall 2B
- Hadas Orgad (Technion), Michael Toker (Technion), Zorik Gekhman (Technion), Roi Reichart (Technion), Idan Szpektor (Google Research), Hadas Kotek, Yonatan Belinkov (Technion)
- POSTER
- Large-Scale Image-Caption Data in Improving Multimodal Foundation Models
- 10:00 - 12:30, #256, Poster Session 3, Hall 3 + Hall 2B
- Jeff Lai, Vasileios Saveris, Chen Chen, Hong-You Chen, Haotian Zhang, Bowen Zhang, Wenze Hu, Juan Lao Tebar, Zhe Gan, Peter Grasch, Meng Cao, Yinfei Yang
- POSTER
- MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning
- 10:00 - 12:30, #223, Poster Session 3, Hall 3 + Hall 2B
- Haotian Zhang, Mingfei Gao, Zhe Gan, Philipp Dufter, Nina Wenzel, Forrest Huang, Dhruti Shah, Xianzhi Du, Bowen Zhang, Yanghao Li, Sam Dodge, Keen You, Zhen Yang, Aleksei Timofeev, Mingze Xu, Hong-You Chen, Jean-Philippe Fauconnier Biard, Jeff Lai, Haoxuan You, Zirui Wang, Afshin Dehghan, Peter Grasch, Yinfei Yang
- POSTER
- MMEgo: Towards Building Egocentric Multimodal LLMs for Video QA
- 10:00 - 12:30, #205, Poster Session 3, Hall 3 + Hall 2B
- Hanrong Ye (Hong Kong University of Science and Technology (HKUST)), Haotian Zhang, Erik Daxberger, Lin Chen, Zongyu Lin (UCLA), Yanghao Li, Bowen Zhang, Haoxuan You, Jiasen Lu, Dan Xu (HKUST), Zhe Gan, Yinfei Yang
- POSTER
- The AdEMAMix Optimizer: Better, Faster, Older
- 10:00 - 12:30, #379, Poster Session 3, Hall 3 + 2B
- Matteo Pagliardini (EPFL), Pierre Ablin, David Grangier
- POSTER
- Theory, Analysis, and Best Practices for Sigmoid Self-Attention
- 10:00 - 12:30, #596, Poster Session 3, Hall 3 + 2B
- Jason Ramapuram, Federico Danieli, Eeshan Gunesh Dhekane, Floris Weers, Dan Busbridge, Pierre Ablin, Tatiana Likhomanenko, Jagrit Digani, Zijin Gu, Amitis Shidani, Russ Webb
- POSTER
- TIS-DPO: Token-level Importance Sampling for Direct Preference Optimization
- 10:00 - 12:30, #540, Poster Session 3, Hall 3 + Hall 2B
- Law Liu (Tsinghua University), Felix Bai, Zhiyun Lu, Yanchao Sun, Xiang Kong, Simon Wang, Jiulong Shan, Lijie Wen (Tsinghua University), Philip S. Yu (University of Illinois at Chicago), Meng Cao
- SOCIAL
- Women in Machine Learning (WiML)
- 12:30 - 14:00, Conference GHJ
- Helen Zhou and Nandita Bhaskhar will represent Apple at the WiML social.
- POSTER
- Do LLMs Know Internally When They Follow Instructions?
- 15:00 - 17:30, #534, Poster Session 4, Hall 3 + Hall 2B
- Juyeon Heo (University of Cambridge), Christina Heinze-Deml, Oussama Elachqar, Shirley Ren, Udhay Nallasamy, Andy Miller, Kwan Ho Ryan Chan (University of Pennsylvania), Jaya Narain
- POSTER
- Does Spatial Cognition Emerge in Frontier Models?
- 15:00 - 17:30, #251, Poster Session 4, Hall 3 + Hall 2B
- Santhosh Kumar Ramakrishnan, Erik Wijmans, Philipp Krähenbühl, Vladlen Koltun
- POSTER
- Efficient Source-Free Time-Series Adaptation via Parameter Subspace Disentanglement
- 15:00 - 17:30, #351, Poster Session 4, Hall 3 + Hall 2B
- Gaurav Patel (Purdue University), Chris Sandino, Behrooz Mahasseni, Ellen Zippi, Erdrin Azemi, Ali Moin, Juri Minxha
- POSTER
- GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models
- 15:00 - 17:30, #223, Poster Session 4, Hall 3 + Hall 2B
- Iman Mirzadeh, Keivan Alizadeh Vahid, Hooman Shahrokhi (Washington State University), Oncel Tuzel, Samy Bengio, Mehrdad Farajtabar
- POSTER
- MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs
- 15:00 - 17:30, #218, Poster Session 4, Hall 3 + Hall 2B
- Yusu Qian, Hanrong Ye (HKUST), Jean-Philippe Fauconnier Biard, Peter Grasch, Yinfei Yang, Zhe Gan
- ORAL PRESENTATION
- Cut Your Losses in Large-Vocabulary Language Models
- 15:30 - 15:42, Peridot 204-205
- Erik Wijmans, Brody Huval, Alexander Hertzberg, Vladlen Koltun, Philipp Krähenbühl
- SOCIAL
- LatinX in AI
- 17:00 - 18:30, Conference GHJ
- Alejandro Newell and Miguel Sarabia del Castillo will represent Apple at the LatinX in AI social.
Saturday, April 26
- POSTER
- A Formal Framework for Understanding Length Generalization in Transformers
- 10:00 - 12:30, #457, Poster Session 5, Hall 3 + Hall 2B
- Xinting Huang (Saarland University), Andy Yang (University of Notre Dame), Yash Sarrof (Saarland University), Mark Rofin (Saarland University), Satwik Bhattamishra (University of Oxford), Andreas Krebs (University of Tübingen), Hattie Zhou (MILA), Preetum Nakkiran, Michael Hahn (Saarland University)
- POSTER
- Disentangled Representational Learning with the Gromov-Monge Gap
- 10:00 - 12:30, #603, Poster Session 5, Hall 3 + Hall 2B
- Théo Uscidda (CREST-ENSAE), Lucas Eyring (TUM), Karsten Roth (Tübingen), Fabian Theis (TUM), Zeynep Akata (TUM), Marco Cuturi
- POSTER
- Talking Turns: Benchmarking Audio Foundation Models on Turn-Taking Dynamics
- 10:00 - 12:30, #50, Poster Session 5, Hall 3 + Hall 2B
- Siddhant Arora (CMU), Zhiyun Lu, Chung-Cheng Chiu, Ruoming Pang, Shinji Watanabe (CMU)
- POSTER
- Scaling Diffusion Language Models via Adaptation from Autoregressive Models
- 10:00 - 12:30, #609, Poster Session 5, Hall 3 + Hall 2B
- Shansan Gong (HKU), Shivam Agarwal (UIUC), Yizhe Zhang, Lin Zheng (HKU), Jiacheng Ye (HKU), Mukai Li (HKU), Chenxin An (HKU), Hao Peng (UIUC), Lingpeng Kong (HKU)
- POSTER
- Task-Adaptive Pretrained Language Models via Clustered-Importance Sampling
- 10:00 - 12:30, #286, Poster Session 5, Hall 3 + Hall 2B
- David Grangier, Simin Fan (EPFL), Skyler Seto, Pierre Ablin
- POSTER
- CoMotion: Concurrent Multi-person 3D Motion Through Time
- 15:00 - 17:30, #127, Poster Session 6, Hall 3 + Hall 2B
- Alejandro Newell, Peiyun Hu, Lahav Lipson, Stephan Richter, Vladlen Koltun
- POSTER
- DART: Denoising Autoregressive Transformer for Scalable Text-to-Image Generation
- 15:00 - 17:30, #625, Poster Session 4, Hall 3 + Hall 2B
- Jiatao Gu, Shuangfei Zhai, Yuyang Wang, Qihang Zhang (The Chinese University of Hong Kong), Yizhe Zhang, Dinghuai Zhang (Mila), Navdeep Jaitly, Josh Susskind
- POSTER
- Depth Pro: Sharp Monocular Metric Depth in Less Than a Second
- 15:00 - 17:30, #115, Poster Session 6, Hall 3 + Hall 2B
- Aleksei Bochkovskii, Amael Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan Richter, Vladlen Koltun
- POSTER
- RelCon: Relative Contrastive Learning for a Motion Foundation Model for Wearable Data
- 15:00 - 17:30, #17, Poster Session 6, Hall 3 + Hall 2B
- Max Xu (UIUC), Jaya Narain, Greg Darnell, Hyewon Jeong (MIT), Haraldur Hallgrimsson, Darren Forde, Richard Fineman, James M. Rehg (UIUC), Karthik Jayaraman Raghuram, Shirley Ren
- POSTER
- Simple ReFlow: Improved Techniques for Fast Flow Models
- 15:00 - 17:30, #581, Poster Session 6, Hall 3 + Hall 2B
- Beomsu Kim, Yu-Guan Hsieh, Bahjat Kawar, Marco Cuturi, Jong Chul Ye (KAIST), James Thornton, Michal Klein
- SOCIAL
- Queer in AI
- 17:00 - 18:30, Conference GHJ
- Azim Yusoff, Kevin Miao, and Nate True will represent Apple at the Queer in AI social.
Sunday, April 27
- WORKSHOP
- Foundation Models in the Wild 2025
- 08:00 – 17:15, Hall 4, #6
-
- POSTER
- FocalLens: Instruction Tuning Enables Zero-Shot Conditional Image Representations
- Cheng-Yu Hsieh (University of Washington), Pavan Kumar Anasosalu Vasu, Fartash Faghri, Raviteja Vemulapalli, Chun-Liang Li, Ranjay Krishna, Oncel Tuzel, Hadi Pour Ansari
- WORKSHOP
- Sparsity in LLMs (SLLM): Deep Dive into Mixture of Experts, Quantization, Hardware, and Inference
- 09:00 - 18:00, Hall 4, #7
-
- POSTER
- From Dense to Dynamic: Token-Difficulty Driven MoEfication of Pre-Trained LLMs
- Kumari Nishu, Sachin Mehta, Samira Abnar, Mehrdad Farajtabar, Max Horton, Mahyar Najibi, Moin Nabi, Minsik Cho, Devang Naik
- ORAL PRESENTATION AND POSTER
- Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for Mixture-of-Experts Language Models
- Samira Abnar, Harshay Shah (MIT), Dan Busbridge, Alaaeldin El-Nouby, Josh Susskind, Vimal Thilak
Monday, April 28
- WORKSHOP
- Scalable Optimization for Efficient and Adaptive Foundation Models (SCOPE) 2025
- 08:15 – 17:30, Hall 4, #7
-
- POSTER
- KV Prediction for Improved Time to First Token
- Max Horton, Qingqing Cao, Frank Sun, Yanzi Jin, Sachin Mehta, Mohammad Rastegari, Moin Nabi
- ORAL
- M2R2: Mixture of Multi-Rate Residuals for Efficient Transformer Inference
- Nikhil Bhendawade, Mahyar Najibi, Irina Belousova, Devang Naik
- WORKSHOP
- I Can't Believe It's Not Better: Challenges in Applied Deep Learning (ICBINB) 2025
- 09:00 - 18:00, Hall 4, #1
-
- POSTER
- Modeling Speech Emotion With Label Variance and Analyzing Performance Across Speakers and Unseen Acoustic Conditions
- Vikram Mitra, Tien Dung Tran, Amrit Romana, Erdrin Azemi
Technical Demos
Visit Apple's booth at Singapore EXPO, Booth C03, to see our technical demos during exhibition hours:
- DEMO
- FastVLM
- FastVLM is a family of mobile-friendly vision language models. These on-device models use a mix of CNN and transformer encoding techniques. Designed specifically for on-device applications like chatbots, captions, and image finders. Together, they optimize the balance between accuracy and speed.
- DEMO
- Depth Pro
- Zero-shot monocular depth estimation from images without needing to know anything about the camera during training. Depth Pro can generalize to a wide variety of images including in-the-wild internet photos, low-light, text, and motion-blurred images from a smartphone. It uses a query-based architecture to offer state-of-the-art vision transformer modeling, and it works with both RGB and depth at multiple scales. Results show that Depth Pro has unmatched capability in out-of-domain generalization and accuracy, and it works with all kinds of photos. Absolute depth cues in each local region are provided.
Accepted Papers
- Adaptive Batch Size for Privately Finding Second-order Stationary Points
- Daogao Liu (University of Washington), Kunal Talwar
- The AdEMAMix Optimizer: Better, Faster, Older
- Matteo Pagliardini (EPFL), Pierre Ablin, David Grangier
- CoMotion: Concurrent Multi-person 3D Motion Through Time
- Alejandro Newell, Peiyun Hu, Lahav Lipson, Stephan Richter, Vladlen Koltun
- Controlling Language and Diffusion Models by Transporting Activations
- Pau Rodríguez, Arno Blaas, Michal Klein, Luca Zappella, Nicholas Apostoloff, Marco Cuturi, Xavier Suau
- Cut Your Losses in Large-Vocabulary Language Models
- Erik Wijmans, Brody Huval, Alexander Hertzberg, Vladlen Koltun, Philipp Krähenbühl
- DART: Denoising Autoregressive Transformer for Scalable Text-to-Image Generation
- Jiatao Gu, Shuangfei Zhai, Yuyang Wang, Qihang Zhang (The Chinese University of Hong Kong), Yizhe Zhang, Dinghuai Zhang (Mila), Navdeep Jaitly, Josh Susskind
- Depth Pro: Sharp Monocular Metric Depth in Less Than a Second
- Aleksei Bochkovskii, Amael Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan Richter, Vladlen Koltun
- Disentangled Representational Learning with the Gromov-Monge Gap
- Théo Uscidda (CREST-ENSAE), Lucas Eyring (TUM), Karsten Roth (Tübingen), Fabian Theis (TUM), Zeynep Akata (TUM), Marco Cuturi
- Do LLMs Estimate Uncertainty Well in Instruction-Following?
- Juyeon Heo (University of Cambridge), Miao Xiong (National University of Singapore), Christina Heinze-Deml, Jaya Narain
- Do LLMs Know Internally When They Follow Instructions?
- Juyeon Heo (University of Cambridge), Christina Heinze-Deml, Oussama Elachqar, Shirley Ren, Udhay Nallasamy, Andy Miller, Kwan Ho Ryan Chan (University of Pennsylvania), Jaya Narain
- Does Spatial Cognition Emerge in Frontier Models?
- Santhosh Kumar Ramakrishnan, Erik Wijmans, Philipp Krähenbühl, Vladlen Koltun
- EC-DiT: Scaling Diffusion Transformers With Expert Choice Routing
- Haotian Sun (Georgia Institute of Technology), Tao Lei, Bowen Zhang, Yanghao Li, Haoshuo Huang, Ruoming Pang, Bo Dai (Georgia Institute of Technology), Nan Du
- Efficient Source-Free Time-Series Adaptation via Parameter Subspace Disentanglement
- Gaurav Patel (Purdue University), Chris Sandino, Behrooz Mahasseni, Ellen Zippi, Erdrin Azemi, Ali Moin, Juri Minxha
- Ferret-UI 2: Mastering Universal User Interface Understanding Across Platforms
- Zhangheng Li, Keen You, Haotian (AIML) Zhang, Di Feng, Harsh Agrawal, Xiujun Li, Mohana Prasad Sathya Moorthy, Jeff Nichols, Yinfei Yang, Zhe Gan
- A Formal Framework for Understanding Length Generalization in Transformers
- Xinting Huang (Saarland University), Andy Yang (University of Notre Dame), Yash Sarrof (Saarland University), Mark Rofin (Saarland University), Satwik Bhattamishra (University of Oxford), Andreas Krebs (University of Tübingen), Hattie Zhou (MILA), Preetum Nakkiran, Michael Hahn (Saarland University)
- GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models
- Iman Mirzadeh, Keivan Alizadeh Vahid, Hooman Shahrokhi (Washington State University), Oncel Tuzel, Samy Bengio, Mehrdad Farajtabar
- How to Verify Any (Reasonable) Distribution Property: Computationally Sound Argument Systems for Distributions
- Tal Herman (Weizmann Institute of Science), Guy Rothblum
- Language Models Know More Than They Show: Exploring Hallucinations From the Model's Viewpoint
- Hadas Orgad (Technion), Michael Toker (Technion), Zorik Gekhman (Technion), Roi Reichart (Technion), Idan Szpektor (Google Research), Hadas Kotek, Yonatan Belinkov (Technion)
- Large-Scale Image-Caption Data in Improving Multimodal Foundation Models
- Jeff Lai, Vasileios Saveris, Chen Chen, Hong-You Chen, Haotian Zhang, Bowen Zhang, Wenze Hu, Juan Lao Tebar, Zhe Gan, Peter Grasch, Meng Cao, Yinfei Yang
- MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs
- Yusu Qian, Hanrong Ye (HKUST), Jean-Philippe Fauconnier Biard, Peter Grasch, Yinfei Yang, Zhe Gan
- MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning
- Haotian Zhang, Mingfei Gao, Zhe Gan, Philipp Dufter, Nina Wenzel, Forrest Huang, Dhruti Shah, Xianzhi Du, Bowen Zhang, Yanghao Li, Sam Dodge, Keen You, Zhen Yang, Aleksei Timofeev, Mingze Xu, Hong-You Chen, Jean-Philippe Fauconnier Biard, Jeff Lai, Haoxuan You, Zirui Wang, Afshin Dehghan, Peter Grasch, Yinfei Yang
- MMEgo: Towards Building Egocentric Multimodal LLMs for Video QA
- Hanrong Ye (Hong Kong University of Science and Technology (HKUST)), Haotian Zhang, Erik Daxberger, Lin Chen, Zongyu Lin (UCLA), Yanghao Li, Bowen Zhang, Haoxuan You, Jiasen Lu, Dan Xu (HKUST), Zhe Gan, Yinfei Yang
- No Need to Talk: Training Mixture of Language Models Independently
- Anastasiia Filippova (EPFL), Ronan Collobert, Angelos Katharopoulos, David Grangier
- On the Modeling Capabilities of Large Language Models for Sequential Decision Making
- Martin Klissarov (McGill University / Mila), Devon Hjelm, Alexander Toshev, Bogdan Mazoure
- Provable Uncertainty Decomposition via Higher-Order Calibration
- Gustaf Ahdritz (Harvard University), Aravind Gollakota, Parikshit Gopalan, Charlotte Peale (Stanford University), Udi Wieder
- RelCon: Relative Contrastive Learning for a Motion Foundation Model for Wearable Data
- Max Xu (UIUC), Jaya Narain, Greg Darnell, Hyewon Jeong (MIT), Haraldur Hallgrimsson, Darren Forde, Richard Fineman, James M. Rehg (UIUC), Karthik Jayaraman Raghuram, Shirley Ren
- Scaling Diffusion Language Models via Adaptation From Autoregressive Models
- Shansan Gong (HKU), Shivam Agarwal (UIUC), Yizhe Zhang, Lin Zheng (HKU), Jiacheng Ye (HKU), Mukai Li (HKU), Chenxin An (HKU), Hao Peng (UIUC), Lingpeng Kong (HKU)
- SeedLM: Compressing LLM Weights through Seeds of a Pseudo-Random Generator
- Rasoul Shafipour, David Harrison, Max Horton, Jeff Marker, Houman Bedayat, Sachin Mehta (Meta), Mohammad Rastegari (Meta), Mahyar Najibi, Saman Naderiparizi
- Simple ReFlow: Improved Techniques for Fast Flow Models
- Beomsu Kim, Yu-Guan Hsieh, Bahjat Kawar, Marco Cuturi, Jong Chul Ye (KAIST), James Thornton, Michal Klein
- Step-by-Step Reasoning for Math Problems via Twisted Sequential Monte Carlo
- Shengyu Feng (CMU), Xiang Kong, Shuang Ma, Aonan Zhang, Dong Yin, Chong Wang, Ruoming Pang, Yiming Yang (CMU)
- Talking Turns: Benchmarking Audio Foundation Models on Turn-Taking Dynamics
- Siddhant Arora (CMU), Zhiyun Lu, Chung-Cheng Chiu, Ruoming Pang, Shinji Watanabe (CMU)
- Task-Adaptive Pretrained Language Models via Clustered-Importance Sampling
- David Grangier, Simin Fan (EPFL), Skyler Seto, Pierre Ablin
- Theory, Analysis, and Best Practices for Sigmoid Self-Attention
- Jason Ramapuram, Federico Danieli, Eeshan Gunesh Dhekane, Floris Weers, Dan Busbridge, Pierre Ablin, Tatiana Likhomanenko, Jagrit Digani, Zijin Gu, Amitis Shidani, Russ Webb
- TIS-DPO: Token-level Importance Sampling for Direct Preference Optimization
- Law Liu (Tsinghua University), Felix Bai, Zhiyun Lu, Yanchao Sun, Xiang Kong, Simon Wang, Jiulong Shan, Lijie Wen (Tsinghua University), Philip S. Yu (University of Illinois at Chicago), Meng Cao
Workshop Accepted Papers
- FocalLens: Instruction Tuning Enables Zero-Shot Conditional Image Representations
- Cheng-Yu Hsieh (University of Washington), Pavan Kumar Anasosalu Vasu, Fartash Faghri, Raviteja Vemulapalli, Chun-Liang Li, Ranjay Krishna, Oncel Tuzel, Hadi Pour Ansari
- From Dense to Dynamic: Token-Difficulty Driven MoEfication of Pre-Trained LLMs
- Kumari Nishu, Sachin Mehta, Samira Abnar, Mehrdad Farajtabar, Max Horton, Mahyar Najibi, Moin Nabi, Minsik Cho, Devang Naik
- KV Prediction for Improved Time to First Token
- Max Horton, Qingqing Cao, Frank Sun, Yanzi Jin, Sachin Mehta, Mohammad Rastegari, Moin Nabi
- M2R2: Mixture of Multi-Rate Residuals for Efficient Transformer Inference
- Nikhil Bhendawade, Mahyar Najibi, Irina Belousova, Devang Naik
- Modeling Speech Emotion With Label Variance and Analyzing Performance Across Speakers and Unseen Acoustic Conditions
- Vikram Mitra, Tien Dung Tran, Amrit Romana, Erdrin Azemi
- Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for Mixture-of-Experts Language Models
- Samira Abnar, Harshay Shah (MIT), Dan Busbridge, Alaaeldin El-Nouby, Josh Susskind, Vimal Thilak
Acknowledgements
Alexander Toshev and Ronan Collobert are Senior Area Chairs.
Chen Huang, Chong Wang, Eugene Ndiaye, Harsh Agrawal, Pau Rodriguez, Preetum Nakkiran, Stephan Richter, Yizhe Zhang, and Zhe Gan are Area Chairs.
Arno Blaas is a Workshop Co-Organizer, and Nicholas Apostoloff and Niv Sivakumar are Workshop Reviewers for "I Can't Believe It's Not Better: Challenges in Deep Learning (ICBINB) 2025".
Agni Kumar, Andrew Szot, Arno Blaas, Barry Theobald, Bhuwan Dhingra, Devon Helm, Fartash Faghri, Hadi Pour Ansari, Haoxuan You, Huangjie Zheng, Iman Mirzadeh, Juri Minxha, Kunal Talwar, Lin Chen, Louis Bethune, Luca Zappella, Maartje ter Hoeve, Max Horton, Michael Kirchhof, Nicholas Apostoloff, Pavan Kumar Anasosalu Vasu, Philipp Krähenbühl, Pierre Ablin, Rasoul Shafipour, Raviteja Vemulapalli, Rin Metcalf Susa, Ruochen (Esther) Zhao, Santhosh Kumar Ramakrishnan, Vimal Thilak, Xavier Suau Cuadros, Xiaoming Zhao, and Zakhar Shumaylov are Reviewers.
Related readings and updates.
Apple is presenting new research at the European Conference on Computer Vision (ECCV), which takes place in person in Milan, Italy, from September 29 - October 4. We are proud to again sponsor the biennial conference, which brings together the scientific and industrial research communities around ML and computer vision. Below is an overview of Apple’s participation at ECCV 2024.
Apple sponsored the International Conference on Learning Representations (ICLR), which took place in person from May 7 to 11 in Vienna Austria. ICLR brings together professionals dedicated to the advancement of deep learning.
Discover opportunities in Machine Learning.
Our research in machine learning breaks new ground every day.