AVID learns multi-stage tasks via CycleGAN pixel-level translation of human videos, requiring only 20 minutes of demonstrations.

Laura Smith, Nikita Dhawan, Marvin Zhang et al.

DeepMUSIC employs multiple CNNs on subdivided angular spectra for multi-target DOA estimation, outperforming traditional methods in accuracy and efficiency.

Ahmet M. Elbir

Learn 6DoF closed-loop grasping from low-cost demonstrations using action-view rendering and Q-function for high success rates.

Shuran Song, Andy Zeng, Johnny Lee et al.

DeepDeform learns non-rigid RGB-D reconstruction from semi-supervised data: 390k frames, 5,533 pairs, and state-of-the-art accuracy.

Aljaž Božič, Michael Zollhöfer, Christian Theobalt et al.

Proposes LiDAR Iris, a global descriptor combining LoG-Gabor filtering and Fourier transform for fast, rotation-invariant loop closure detection.

Ying Wang, Zezhou Sun, Cheng-Zhong Xu et al.

Introduced Localized Narratives method, annotating 849k images by linking vision and language.

Jordi Pont-Tuset, Jasper Uijlings, Soravit Changpinyo et al.

Introduces CLOTH3D dataset and GCVAE model for realistic 3D clothed human generation, capturing garment topology and dynamics.

Hugo Bertiche, Meysam Madadi, Sergio Escalera

AUGMIX combines stochastic augmentation and Jensen-Shannon loss to improve image classifier robustness and uncertainty, reducing corruption errors by over 50% on benchmarks.

Dan Hendrycks, Norman Mu, Ekin D. Cubuk et al.

Introduces ‘Effective Model Complexity’ to explain double descent phenomena, revealing non-monotonic effects of model size and training epochs on generalization.

Preetum Nakkiran, Gal Kaplun, Yamini Bansal et al.

StarGAN v2 achieves diverse image synthesis across multiple domains, significantly outperforming baselines.

Yunjey Choi, Youngjung Uh, Jaejun Yoo et al.

ALFRED benchmark uses 25k natural language directives and expert demonstrations to challenge embodied vision-language models in complex household tasks.

Mohit Shridhar, Jesse Thomason, Daniel Gordon et al.

Dreamer leverages latent space imagination with deep models and analytic gradients to achieve efficient long-horizon control on 20 visual tasks, surpassing prior methods.

Danijar Hafner, Timothy Lillicrap, Jimmy Ba et al.

Redesigning normalization and removing progressive growing, the improved StyleGAN reduces artifacts, enhances image quality, and boosts invertibility, enabling higher resolution outputs.

Tero Karras, Samuli Laine, Miika Aittala et al.

TutorialVQA dataset for instructional video QA, including 6,000 manually collected triples.

Anthony Colas, Seokhwan Kim, Franck Dernoncourt et al.

Proposes Continuous Graph Neural Networks (CGNN) using neural ODEs, addressing over-smoothing and capturing long-range dependencies, achieving state-of-the-art node classification accuracy.

Louis-Pascal A. C. Xhonneux, Meng Qu, Jian Tang

Deep learning seq2seq models achieve over 95% accuracy in symbolic integration and differential equations, outperforming Mathematica and Matlab.

Guillaume Lample, François Charton

Automatic prompt generation using relation mining and paraphrasing improves LM knowledge extraction accuracy from 31.1% to 39.6%.

Zhengbao Jiang, Frank F. Xu, Jun Araki et al.

Introduces SAMSum corpus with manually annotated abstractive dialogue summaries; models achieve higher ROUGE scores on dialogues than news, but human judgment contradicts.

Bogdan Gliwa, Iwona Mochol, Maciej Biesek et al.

Constructed ATCSpeech, a real multi-language ATC speech corpus, and trained deep learning ASR models achieving CER of 12-15%, enhancing aviation safety.

Bo Yang, Xianlong Tan, Zhengmao Chen et al.

PQ-NET generates 3D shapes via sequential part assembly, enabling autoencoding, interpolation, and single-view reconstruction.

Rundi Wu, Yixin Zhuang, Kai Xu et al.