Publications
Full chronological list. For an authoritative, continuously-updated view with citation counts, see Google Scholar →.
2026
- Multiplicative vs. additive state coupling in learned optimizer schedulesNeurIPS 2026 OPT-ML Workshop
- Graph coarsening enables escaping local optima in compute-constrained nonconvex MAX-CUTNeurIPS 2026 OPT Workshop
- Alleviating hallucination in reasoning tasks with training-free uncertainty-guided steeringNeural Information Processing Systems (NeurIPS 2026)
- On locality and length generalization in visual reasoningEuropean Conference on Computer Vision (ECCV 2026)
- Streaming interventions: Can video large language models correct mistakes as they occur?Neural Information Processing Systems (NeurIPS 2026)
- Notes-to-Self: Scratchpad augmented VLAs for memory-dependent manipulation tasksInternational Conference on Robotics and Automation (ICRA 2026)
- On the “induction bias” in sequence modelsInternational Conference on Machine Learning (ICML 2026)
- From out-of-distribution detection to hallucination detection: A geometric viewInternational Conference on Machine Learning (ICML 2026)
- Can vision-language models answer face-to-face questions in the real world?International Conference on Learning Representations (ICLR 2026)
- Enhancing hallucination detection through noise injectionInternational Conference on Learning Representations (ICLR 2026)
2025
- Can multi-modal LLMs provide live step-by-step task guidance?Neural Information Processing Systems (NeurIPS 2025)
- Revisiting bi-linear state transitions in recurrent neural networksNeural Information Processing Systems (NeurIPS 2025)
- Multi-draft speculative sampling: Canonical decomposition and theoretical limitsInternational Conference on Learning Representations (ICLR 2025)
- Delayed attention training improves length generalization in transformer–RNN hybridsNeurIPS 2025 Workshop on What Transformers Can(not) Do
- Replacing thinking with tool usage enables reasoning in small language modelsarXiv:2507.05065
- Grounding code understanding in step-by-step executionInternational Conference on Learning Representations (ICLR 2025)
2024
- ClevrSkills: Compositional language and visual reasoning in roboticsNeural Information Processing Systems (NeurIPS 2024)
- Your context is not an array: Unveiling random access limitations in transformersConference on Language Modeling (COLM 2024)
- AirLetters: An open video dataset of characters drawn in the airEuropean Conference on Computer Vision (ECCV 2024), HANDS Workshop
- Look, remember and reason: Grounded reasoning in videos with language modelsInternational Conference on Learning Representations (ICLR 2024)
- Towards bridging classical and neural computation through a read-eval-print loopICML 2024 Workshop on LLMs and Cognition
- PlotTwist: Vision-language models struggle with reasoning over mathematical plotsNeurIPS 2024 EVAL-FoMo Workshop (Evaluating Foundation Models)
- Math-Search: A benchmark for multi-hop visual reasoning over plotsCVPR 2024 MAR Workshop (Multimodal Algorithmic Reasoning) · earlier version of the PlotTwist work
2023
- Deductive verification of chain-of-thought reasoningNeural Information Processing Systems (NeurIPS 2023)
- Unleashing the creative mind: Language model as hierarchical policy for improved exploration on challenging problem solvingarXiv:2311.00694
2022
- Decision making as language generationNeurIPS Workshop on Foundation Models for Decision Making
2020
- Putting the skeleton back in the closetMedium (TwentyBN blog) · a later version of this work appeared as “Is end-to-end learning enough for fitness activity recognition?” (2023)
2019
- The Jester dataset: A large-scale video dataset of human gesturesInternational Conference on Computer Vision (ICCV 2019), HANDS Workshop
2018
- The more fine-grained, the better for transfer learningNeurIPS 2018 Workshop on Modeling and Decision-Making in the Spatiotemporal Domain
- Evaluating visual “common sense” using fine-grained classification and captioning tasksICLR 2018 Workshop
2017
- The “something something” video database for learning and evaluating visual common senseInternational Conference on Computer Vision (ICCV 2017)
- Incorporating long-range consistency in CNN-based texture generationInternational Conference on Learning Representations (ICLR 2017)
- Denoising criterion for variational auto-encoding frameworkAAAI Conference on Artificial Intelligence (AAAI 2017)
2016
- Neural networks with few multiplicationsInternational Conference on Learning Representations (ICLR 2016)
- Regularizing RNNs by stabilizing activationsInternational Conference on Learning Representations (ICLR 2016)
- Conservativeness of untied auto-encodersAAAI Conference on Artificial Intelligence (AAAI 2016)
- Architectural complexity measures of recurrent neural networksNeural Information Processing Systems (NIPS 2016)
- EmoNets: Multimodal deep learning approaches for emotion recognition in videoJournal on Multimodal User Interfaces, 10(2) · built on the EmotiW 2013 Grand Challenge-winning system
- Generative adversarial metricICLR 2016 Workshop
2015
- The potential energy of an autoencoderIEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI)
- Zero-bias autoencoders and the benefits of co-adapting featuresInternational Conference on Learning Representations (ICLR 2015)
- Learning visual odometry with a convolutional networkInternational Conference on Computer Vision Theory and Applications (VISAPP 2015)
- On using very large target vocabulary for neural machine translationAnnual Meeting of the Association for Computational Linguistics (ACL 2015)
- Real-time activity recognition via deep learning of motion featuresEuropean Symposium on Artificial Neural Networks (ESANN 2015)
- Recurrent neural networks for emotion recognition in videoACM International Conference on Multimodal Interaction (ICMI 2015)
- Montreal neural machine translation systems for WMT'15Tenth Workshop on Statistical Machine Translation (WMT 2015)
2014
- Modeling deep temporal dependencies with recurrent “grammar cells”Neural Information Processing Systems (NIPS 2014)
- Learning to encode motion using spatio-temporal synchronyInternational Conference on Learning Representations (ICLR 2014)
- A unified approach to learning depth and motion featuresIndian Conference on Computer Vision, Graphics and Image Processing (ICVGIP 2014)
2013
- Learning invariant features by harnessing the aperture problemInternational Conference on Machine Learning (ICML 2013)
- Feature grouping from spatially constrained multiplicative interactionInternational Conference on Learning Representations (ICLR 2013)
- Combining modality-specific deep neural networks for emotion recognition in videoICMI 2013 — winner, EmotiW Grand Challenge
2012
- Shared kernel information embedding for discriminative inferenceIEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI)
- On spatio-temporal feature learningThe Learning Workshop, Snowbird
2011
- Gradient-based learning of higher-order image featuresInternational Conference on Computer Vision (ICCV 2011)
- Modeling the joint density of two images under a variety of transformationsComputer Vision and Pattern Recognition (CVPR 2011)
- On spatio-temporal sparse coding: Analysis and an algorithmNIPS Workshop on Deep Learning and Unsupervised Feature Learning
2010
- Learning to represent spatial transformations with factored higher-order Boltzmann machinesNeural Computation, 22(6): 1473–1492 · cover article
2009
- Shared kernel information embedding for discriminative inferenceComputer Vision and Pattern Recognition (CVPR 2009)
2007
- Unsupervised learning of image transformationsComputer Vision and Pattern Recognition (CVPR 2007)
2006
- Propagating errors and beliefs for large scale nonlinear structure predictionNorth East Student Colloquium on Artificial Intelligence (NESCAI), Ithaca, NY
2005
- Principal surfaces from unsupervised kernel regressionIEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 27(9): 1379–1391
- Embedding via clustering: Using spectral information to guide dimensionality reductionInternational Joint Conference on Neural Networks (IJCNN 2005)
- Dual optimization of conditional probability modelsNIPS Workshop on Kernel Methods for Structured Domains
- Exploring the regularization path for adaptive Gaussian kernel SVMsNIPS Workshop on the Accuracy–Regularization Frontier
Theses
- Non-linear latent factor models for revealing structure in high-dimensional dataPhD thesis, Department of Computer Science, University of Toronto, 2008 · advisor: Geoffrey Hinton
- Unsupervised kernel regression for nonlinear dimensionality reductionMaster's thesis, Department of Computer Science, Bielefeld University, 2003 · advisors: Helge Ritter and Peter Meinicke
See citation counts and additional papers on Google Scholar →