Email
bengio [at] apple.com
Location
Apple Park, Cupertino, CA, USA
Samy Bengio
Senior Director, AI & ML Research at Apple Machine Learning Research
Adjunct Professor at EPFL
I lead Artificial Intelligence and Machine Learning Research at Apple. My research spans deep learning foundations, the limits of reasoning in large language and autoregressive models, multimodal intelligence, and next-generation machine learning systems.
Over the course of my career, I was involved in the inception and early development of seminal open-source ML libraries including Torch, SVMTorch, and TensorFlow. More recently at Apple, I have been sponsoring our open-source MLX framework for efficient machine learning on Apple Silicon.
Publications & Vitae
Over three decades of research contributions across machine learning theory, vision, speech, and systems.
Complete Publications Archive
Full catalog of peer-reviewed conference and journal publications spanning 1990 to 2026.
Browse Full List →Google Scholar Profile
Track citations, metrics, h-index, and co-authors across all published work.
View Google Scholar →Curriculum Vitae
Academic positions, industrial leadership, education, managed grants, and duties.
EPFL Lecture Series
Course notes and slides for Statistical Machine Learning from Data taught at EPFL in 2006.
View Course Lectures →Research Focus
Core themes and frontier inquiries spanning theoretical foundations and practical intelligence systems.
Limits of Reasoning in Autoregressive Models
Investigating the fundamental reasoning, planning, and length generalization capabilities of Transformers and autoregressive architectures. Analyzing theoretical globality barriers, inductive scratchpads, and failure modes in agentic settings.
Machine Learning Systems & Libraries
Pioneering the architectures and computational foundations of deep learning software. From early building blocks in Torch and SVMTorch, to large-scale distributed training in TensorFlow, and Apple Silicon native acceleration in MLX.
Understanding Training & Generalization
Theoretical and empirical investigations into why deep overparameterized networks generalize well, the role of optimization algorithms, implicit regularization, and unsupervised representation learning.
Deep Architectures for Sequences & Structures
Modeling complex sequential and structured data (speech, natural language, multimodal streams) with deep architectures, addressing rare word representations, curriculum learning, and long-tail distributions.
Adversarial Robustness & Training
Analyzing the vulnerability of neural networks to adversarial attacks in both digital and physical domains, and developing robust adversarial training schemes at scale.
Multimodal Learning & Image Captioning
Landmark explorations in cross-modal understanding, translating perceptual visual inputs into natural descriptive language via end-to-end neural sequence generation.
Invited Talks & Keynotes
Keynotes, invited lectures, and tutorials presented at international conferences and academic institutions.
ICML 2026 workshop on compositional learning
Keynote talk on "Reasoning with LLMs: Challenges and Opportunities", Seoul, Korea, July 2026.
ICML 2026 workshop on planning in the era of LLMs (LM4Plan)
Keynote talk on "Reasoning with LLMs: Challenges and Opportunities", Seoul, Korea, July 2026.
ICML 2026 workshop on failure modes in agentic AI
Keynote talk on "Reasoning with LLMs: Challenges and Opportunities", Seoul, Korea, July 2026.
Apple workshop on Reasoning and Planning'2026
Keynote talk on "Reasoning with LLMs: Challenges and Opportunities", Cupertino, USA, June 2026.
Kempner Institute at Harvard
Keynote talk on "Reasoning with LLMs: Challenges and Opportunities", Boston, USA, May 2026.
ICLR 2026 Workshop on Latent and Implicit Thinking Going Beyond CoT Reasoning
Keynote talk on "Reasoning with LLMs: Challenges and Opportunities", Rio de Janeiro, Brazil, April 2026.
Summit of AI in LATAM (SALA) 2026
Keynote talk on "Reasoning with LLMs: Challenges and Opportunities", Quito, Ecuador, March 2026.
Cornell CS Colloquium 2025
Keynote talk on "Recent results towards reasoning with LLMs", Ithaca, NY, USA, September 2025.
EEML 2025
Keynote talk on "Recent results towards reasoning with LLMs", Sarajevo, July 2025.
Apple workshop on Reasoning and Planning'2025
Keynote talk on "Recent results towards reasoning", Cupertino, July 2025.
IJCNN 2025
Keynote talk on "Recent results towards reasoning with LLMs", Rome, July 2025.
Apple workshop on Natural Language and Interactive Systems'2025
Keynote talk on "Recent results towards reasoning", Cupertino, May 2025.
IDIAP Keynote 2025
Keynote talk on "How far can transformers reason? the globality barrier and inductive scratchpad", Martigny, February 2025.
CMU Keynote 2025
Keynote talk on "How far can transformers reason? the globality barrier and inductive scratchpad", Pittsburgh, February 2025.
Math-AI workshop at NeurIPS'2024
Keynote talk on "How far can transformers reason? the globality barrier and inductive scratchpad", Vancouver, December 2024.
Deep Learning Indaba 2024
Keynote talk on "Learning to Reason is Hard", Dakar, September 2024.
Understanding LLM Understanding Summer School 2024
Keynote talk on "Learning to Reason is Hard", Montreal, June 2024.
AI: reasoning and mathematics 2024
Keynote talk on "Reasoning and Length Generalization Can Be Hard: Two Insights", EPFL, Lausanne, May 2024.
RIIAA 2024
Keynote talk on "Length Generalization Can Be Hard: Two Insights", Quito, Ecuador, February 2024.
ITA 2024
Keynote talk on "Length Generalization Can Be Hard: Two Insights", San Diego, CA, USA, February 2024.
MoroccoAI 2023 conference
Keynote talk on "Generalization on the unseen", Online, December 2023.
NeurIPS Workshop on Efficient Natural Language and Speech Processing 2023
Keynote talk on "Simple and efficient self-training approaches for speech recognition", NeurIPS, New Orleans, USA, December 2023.
ICML Workshop on Knowledge and Logical Reasoning 2023
Keynote talk on "Generalization on the unseen, logic reasoning and degree curriculum", ICML, Honolulu, USA, July 2023.
Mila TechAide 2023
Keynote talk on "Recent Advances in Machine Learning Research at Apple", Mila TechAide, Montreal, Canada, April 2023.
Khipu 2023
Keynote talk on "Recent Advances in Machine Learning Research at Apple", Khipu, Montevideo, Uruguay, March 2023.
BayLearn 2022
Keynote talk on "Recent Advances in Representation Learning", Genentech, South San Francisco, October 2022.
BIFOLD Inauguration
Keynote talk on "Recent Advances in Representation Learning", Berlin, Germany, September 2022.
Applied Machine Learning Days 2022
Keynote talk on "Can Neural Networks Learn Reasoning?", EPFL, Switzerland, March 2022.
Harvard ML Theory Seminar
Keynote talk on "Can Neural Networks Learn Reasoning?", USA, February 2022.
GTC NVIDIA 2021
Keynote talk on "Can Neural Networks Learn Reasoning; Insights, plus a look at long term research in machine learning", USA, November 2021.
Indaba X Senegal 2021
Keynote talk on "Can Neural Networks Learn Reasoning", Dakar, Senegal, August 2021.
World AI Summit Americas 2020
Keynote talk on "Towards Understanding Why Deep Learning Works", Montreal, Canada, March 2020.
ASRU 2019
Keynote talk on "Towards Better Understanding Generalization in Deep Learning", Sentosa, Singapore, December 2019.
Boston University
Invited talk on "Some empirical observations linking representation and generalization", Boston, USA, February 2019.
Qualcomm Machine Learning Summit
Keynote speaker on "Some empirical observations linking representation and generalization", San Diego, USA, October 2018.
Uber Machine Learning Summit
Keynote speaker on "A Sample of Recent Deep Learning Research at Google Brain", San Francisco, USA, September 2018.
Deep Learning: Theory, Algorithms, and Applications
Invited talk on "Neural Combinatorial Optimization with Reinforcement Learning with application to device placement". Berlin, Germany, June 2017.
Yandex
Guest Lecture on "Deep Architectures for Structured Data". Moscow, Russia, December 2016.
Berkeley University
Guest Lecture on "Two Ideas for Structured Data". Berkeley, USA, October 2016.
Deep Learning: Theory, Algorithms, and Applications
Invited talk on "Order Matters: Sequence to Sequence for Sets". MIT, Cambridge, USA, June 2016.
DALI 2016
Invited talk on "Order Matters: Sequence to Sequence for Sets". Sestri Levante, Italy, April 2016.
College de France
Invited talk on "Neural Image Captioning and other Structured Output Problems". Paris, France, March 2016.
eBay Machine Learning and Data Science Conference
Invited talk on "Neural Image Captioning". eBay, USA, March 2016.
Princeton Computer Science Department Seminar
Invited talk on "Neural Image Captioning". Princeton University, USA, February 2016.
NIPS 2015 Extreme Classification Workshop
Invited talk on "Sharing is Caring in the Land of the Long Tail". Montreal, Canada, December 2015.
ICMI 2015
Invited talk on "Sharing Representations for Long Tail Computer Vision Problems ". Seattle, USA, November 2015.
ICIAP 2015
Invited talk on "The Battle Against the Long Tail in Computer Vision". Genoa, Italy, September 2015.
AFEKA Conference for unsupervised learning
Invited talk on "The power of embeddings for text, images, music, voice and image caption generation". Tel Aviv, Israel, April 2015.
Fields Institute for Research in Mathematical Sciences
Invited talk on "The Battle Against the Long Tail". Toronto, Canada, January 2015.
Speaker Odyssey 2014
Invited talk on "Large Scale Learning of a Joint Embedding Space". Joensuu, Finland. June 2014.
Trends in Machine Learning
Invited talk on "Embeddings for images: from large scale image labeling to zero-shot classification". Kyoto, Japan, March 2014.
ERMITES 2013
Invited talks on "Large Scale Semantic Extraction Through Embeddings: From Images to Music" and "DeViSE". Ile de Porquerolles, France, September 2013.
Microsoft Machine Learning Summit 2013
Invited talk on "Laconic: Label Consistency for Image Categorization". Paris, April 2013.
WebVision 2012
Invited talk on "Don't Forget the Long Tail". ECCV 2012, Florence, Italy, October 2012.
Canadian AI 2012
Invited talk on "Learning A Semantic Space: From Image Annotation to Music Similarity". Toronto, Canada. May 2012.
DAS 2012
Conference on Document Analysis Systems. Invited talk on "Learning A Semantic Space: From Image Annotation to Music Similarity". Gold Coast, Queensland. March 2012.
Workshop on Large Scale Multimedia Search 2012
Invited talk on "Learning A Semantic Space: From Image Annotation to Music Similarity". UCLA, January 2011.
NIPS Similarity Workshop 2011
Invited talk on "Online Similarity Learning: From Images to Texts". Sierra Nevada, Andalucia, Spain, December 2011.
Toyota Technical Institute in Chicago
Invited talk on "Learning A Semantic Space: From Image Annotation to Music Similarity". Chicago, November 2011.
National Academy of Engineering's Frontier of Engineering.
Invited talk on "Large Scale Visual Semantic Extraction". Mountain View, California, September 2011.
Symposium on distributed machine learning and sparse representation with massive data sets, organized by CSIRO.
Invited talk on "Web Scale Image Annotation: Learning to Rank with Joint Word-Image Embeddings and Label Embedding Trees". Sydney, Australia, January 2011.
NIPS 2010 Workshop on Machine Learning for Next Generation Computer Vision Challenges.
Invited talk on "Large scale image annotation: Learning to rank with joint word-image embeddings". Whistler, Canada, December 2010.
IEEE Fourth International Conference on Biometrics: Theory, Applications and Systems, BTAS 2010.
Invited talk on "Large scale image annotation: Learning to rank with joint word-image embeddings". Arlington, Virginia, USA, September 2010.
Iberian Conference on Pattern Recognition and Image Analysis, IbPRAI.
Invited talk on "OASIS", Porto, Portugal, June 2009.
Colloque international Francophone sur l'Ecrit et le Document, CIFED.
Invited tutorial on "information extraction: from text to image". Fribourg, September 2006.
ERMITES : Ecole Recherche Multimedia d'Information, Techniques et Sciences.
Lectures on "Statistical Machine Learning for Multimodal Sequence Processing and Extraction". Presqu'ile de Giens, September 2006.
International Conference on Accoustic, Speech and Signal Processing, ICASSP.
Invited talk at special session on Multimodal Biometric Authentication. Title: "Chimeric users to construct fusion classifiers in biometric authentication tasks: An investigation". Toulouse, May 2006.
15eme congres francophone en reconnaissance des formes et intelligence artificielle, RFIA.
Invited talk on "Challenges in Multi Channel Sequence Processing". Tours, January 2006.
International Conference on Multimodal Interactions, ICMI.
Tutorial on "Statistical Machine Learning, with application to Multimodal Interactions". Trento, Italy. October 2005.
Fourteenth Dutch-Belgian Conference on Machine Learning, Benelearn.
Invited talk on "Challenges in Multi-Channel Sequence Processing". University of Twente. Enschede, Holland. February, 2005.
Summer School on Statistics and Environmental Sciences.
Lecture on "Introduction to Statistical Machine Learning". Cargèse, Corsica, France. Novembre, 2004.
Former PhD Students & Postdocs
Mentoring and collaborating with exceptional researchers across academia and industry.
M
Mark Barnard
PhD 2005
F
PhD 2005
S
PhD 2006
R
PhD 2004
C
PhD 2006
N
PhD 2002
D
PhD 2008
M
PhD 2006
J
Johnny Mariéthoz
PhD 2006
M
PhD 2001
J
PhD 2008
N
PhD 2006
A
PhD 2006
Y
PhD 2006
C
postdoc 2004
K
PhD 2003
D
Dong Zhang
PhD 2006
Professional Service & Leadership
Conference leadership, community initiatives, and advisory roles.
🏛️ Major Conference Leadership & Advisory
- I was co ethics review chair for NeurIPS 2021. We wrote this blog post to summarize our processes and findings.
- I co-organized the Foundation of Deep Learning program at the Simons Institute for the Theory of Computing
- I was a member of the scientific advisory board of the Simons Institute for the Theory of Computing
- I was the General Chair of NeurIPS 2018 and the Programme Chair of NIPS 2017
- I have been a Senior Area Chair for all major ML conferences (NeurIPS, ICML and ICLR).
- I am an Action Editor for JMLR.
👥 Workshops & Schools Organized
- The ICML 2019 Workshop on Deep Phenomena
- The 2018 Dagstuhl seminar on Extreme Classification
- The ICML 2017 Reproducibility in Machine Learning Workshop
- The NIPS 2016 Extreme Classification Workshop
- BayLearn, the long standing workshop in machine learning for researchers in the Bay Area.
- The 2016 Bay Area Deep Learning School
- The ICML Extreme Classification Workshop, 2015.
- The AAAI Spring Symposium on Knowledge Representation and Reasoning, 2015.
- BigVision2012, 2014, 2015, a series of workshops on Big Data for Computer Vision.
- WebVision2012, an ECCV 2012 workshop on Computer Vision for the Web.
- The NIPS Workshop on Efficient Machine Learning, Whistler, BC, Canada, December 7-8, 2007.
- The NIPS Workshop on Learning to Compare Examples, Whistler, BC, Canada, December 8-9, 2006.
- The MLMI'06 workshop, Washington DC, USA, May 2006.
- The MLMI'05 workshop, Edinburgh, UK, July 2005.
- The NIPS 2004 Workshop on Multimodal Signal Processing .
- The MLMI'04 workshop, Martigny, Switzerland, June 2004, and the corresponding MLMI'04 recordings of the talks.
- The 2002 IEEE Workshop on Neural Networks for Signal Processing (NNSP), Martigny, Switzerland, September 2002