Antonio Torralba

About

Antonio Torralba is a professor whose research focuses on the development of systems that sense, process, and transmit energy and information, leveraging computational, theoretical, and experimental tools. His work addresses groundbreaking sensors, energy transducers, and physical substrates for computation, aiming to solve shared challenges facing humanity. His expertise spans electrical engineering and computer science, with particular emphasis on areas such as computer vision, machine learning, and artificial intelligence. Throughout his career, Torralba has contributed to advancing the understanding and application of algorithms and systems that interact with the external environment through perception, communication, and action. His research integrates diverse traditions from computer science and electrical engineering to develop techniques for analysis and synthesis of intelligent systems capable of learning, decision-making, and adaptation in changing environments.

Research topics

• Artificial Intelligence
• Computer Science
• Machine Learning
• Human–computer interaction
• Psychology
• Multimedia

Selected publications

• Understanding the role of individual units in a deep neural network

  Proceedings of the National Academy of Sciences · 2020 · 375 citations

  Senior authorCorresponding

  Deep neural networks excel at finding hierarchical representations that solve complex tasks over large datasets. How can we humans understand these learned representations? In this work, we present network dissection, an analytic framework to systematically identify the semantics of individual hidden units within image classification and image generation networks. First, we analyze a convolutional neural network (CNN) trained on scene classification and discover units that match a diverse set of…

  DOI

• WiReSens Toolkit: An Open-source Platform towards Accessible Wireless Tactile Sensing

  2026-03-07

  articleOpen access

  Past research has widely explored the design and fabrication of resistive matrix-based tactile sensors for creating touch-sensitive devices. However, real-world deployment of resistive tactile sensing systems remains difficult for individuals with limited prior experience in embedded sensing due to challenges of portability, adaptivity, and efficiency. We introduce the WiReSens Toolkit, an accessible, open-source platform to bridge this gap. Central to our approach is adaptive hardware for inter…

  PublisherDOI

• MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval

  arXiv (Cornell University) · 2026-04-20

  articleOpen accessSenior author

  Mathematical problem solving remains a challenging test of reasoning for large language and multimodal models, yet existing benchmarks are limited in size, language coverage, and task diversity. We introduce MathNet, a high-quality, large-scale, multimodal, and multilingual dataset of Olympiad-level math problems together with a benchmark for evaluating mathematical reasoning in generative models and mathematical retrieval in embedding-based systems. MathNet spans 47 countries, 17 languages, and…

  PublisherOA PDF

• End-to-End Training for Unified Tokenization and Latent Denoising

  arXiv (Cornell University) · 2026-03-23

  articleOpen access

  Latent diffusion models (LDMs) enable high-fidelity synthesis by operating in learned latent spaces. However, training state-of-the-art LDMs requires complex staging: a tokenizer must be trained first, before the diffusion model can be trained in the frozen latent space. We propose UNITE - an autoencoder architecture for unified tokenization and latent diffusion. UNITE consists of a Generative Encoder that serves as both image tokenizer and latent generator via weight sharing. Our key insight is…

  PublisherOA PDF

• Cooperation by non-kin during birth underpins sperm whale social complexity

  Science · 2026-03-26

  article

  We quantitatively document a sperm whale birth event, revealing collective support behaviors across kinship lines. Using high-resolution drone footage, computer vision, and multiscale network analysis, we studied the interactions within a Caribbean sperm whale unit comprising two matrilines. Our results suggest that a female family member led birth assistance and that after delivery, all individuals oriented toward and helped lift the newborn, taking turns in a coordinated, cross-kin effort. Des…

  PublisherDOI

Recent grants

• CAREER: Integrated system for object and scene recognition

  NSF · $500k · 2008–2013

• NCS-FO: Algorithmically explicit neural representation of visual memorability

  NSF · $805k · 2015–2019

• RI: Small: Advancing Visual Recognition with Feature Visualizations

  NSF · $460k · 2015–2019

Frequent coauthors

• Aude Oliva

  Massachusetts Institute of Technology

  85 shared

• Sanja Fidler

  69 shared

• Joshua B. Tenenbaum

  Massachusetts Institute of Technology

  68 shared

• William T. Freeman

  67 shared

• Carl Vondrick

  48 shared

• David Bau

  48 shared

• Chuang Gan

  45 shared

• Jun-Yan Zhu

  38 shared

Awards & honors

• ACM Fellow (2026)

Similar researchers at Massachusetts Institute of Technology

• Manolis Kellish-164
• Daniela Rush-129
• Anantha Chandrakasanh-115