Proposes Graph Neural Ordinary Differential Equations (GDE), integrating GNNs with ODE solvers for continuous-depth modeling, improving static and dynamic tasks.

Michael Poli, Stefano Massaroli, Junyoung Park et al.

Noisy Student combines large models and noise injection to boost ImageNet accuracy to 88.4%, surpassing previous SOTA with fewer unlabeled images.

Qizhe Xie, Minh-Thang Luong, Eduard Hovy et al.

Introduces kernel mean matching-based instance weighting to reduce distribution shift, significantly improving transfer learning performance.

Fuzhen Zhuang, Zhiyuan Qi, Keyu Duan et al.

Proposes multivariate uncertainty modeling in deep learning using Gaussian loss and Kalman filtering, improving autonomous navigation robustness.

Rebecca L. Russell, Christopher Reale

Relay Policy Learning combines imitation and reinforcement learning with goal-conditioned hierarchical policies to solve long-horizon robotic tasks, outperforming traditional HRL.

Abhishek Gupta, Vikash Kumar, Corey Lynch et al.

HRL4IN employs hierarchical reinforcement learning with multi-space subgoals and embodiment selection, boosting interactive navigation efficiency and energy savings.

Chengshu Li, Fei Xia, Roberto Martin-Martin et al.

Unified T5 model transforms all NLP tasks into text-to-text format, achieving SOTA on benchmarks with up to 11B parameters.

Colin Raffel, Noam Shazeer, Adam Roberts et al.

Introduces multi-pitch features and prototypical triplet loss to improve cover detection accuracy, achieving significant gains on large datasets and live scenarios.

Guillaume Doras, Geoffroy Peeters

Proposed federated neuromorphic SNN training method reduces communication load while maintaining high accuracy, validated on MNIST-DVS with 92.3% accuracy.

Nicolas Skatchkovsky, Hyeryung Jang, Osvaldo Simeone

Proposed ADR algorithm with memory-augmented RL enables sim2real transfer for robot solving Rubik’s cube, achieving 85% success.

OpenAI, Ilge Akkaya, Marcin Andrychowicz et al.

This study analyzes self-play vs. human model learning in cooperation, proposing methods to improve human-AI collaboration using Overcooked experiments.

Micah Carroll, Rohin Shah, Mark K. Ho et al.

Proposes a GCN-GRU deep learning model for dynamic uncertain opinion prediction, addressing scalability, heterogeneity, and conflicting evidence.

Xujiang Zhao, Feng Chen, Jin-Hee Cho

Combines RNN with low-rank Gaussian Copula process for high-dimensional multivariate forecasting, achieving significant accuracy improvements over state-of-the-art methods.

David Salinas, Michael Bohlke-Schneider, Laurent Callot et al.

Checkmate breaks the memory wall with optimal tensor rematerialization, enabling 5.1x larger inputs.

Paras Jain, Ajay Jain, Aniruddha Nrusimha et al.

ZeRO partitions model states to enable trillion-parameter training with reduced memory and high efficiency.

Samyam Rajbhandari, Jeff Rasley, Olatunji Ruwase et al.

Selective-Backprop accelerates deep learning by prioritizing high-loss samples, achieving up to 3.5x faster convergence.

Angela H. Jiang, Daniel L. -K. Wong, Giulio Zhou et al.

Bayesian optimization reveals that low-connectivity (even zero spectral radius) Reservoir networks can effectively forecast chaotic systems.

Aaron Griffith, Andrew Pomerance, Daniel J. Gauthier

DiffTaichi is a differentiable programming language for physical simulation, achieving significant performance improvements.

Yuanming Hu, Luke Anderson, Tzu-Mao Li et al.

CMTS uses variational Bayesian latent space interpolation and style transfer to generate diverse near-miss driving scenarios, improving trajectory prediction robustness.

Wenhao Ding, Mengdi Xu, Ding Zhao

Proposes PCC framework combining prediction, consistency, and low curvature for improved control in high-dimensional observations.

Nir Levine, Yinlam Chow, Rui Shu et al.