Introduces SIMILE, a semantic similarity-based reward for NMT training, improving convergence and translation quality.

John Wieting, Taylor Berg-Kirkpatrick, Kevin Gimpel et al.

Spectral analysis of parameterized weighted Laplacians reveals how parameters influence eigenvalues and clustering stability in large data limits.

Franca Hoffmann, Bamdad Hosseini, Assad A. Oberai et al.

Introduced Jericho environment to study language agents in interactive fiction games, achieving a 10.7% score improvement.

Matthew Hausknecht, Prithviraj Ammanabrolu, Marc-Alexandre Côté et al.

FreiHAND dataset enhances 3D hand pose and shape estimation generalization using multi-view and semi-automated annotation.

Christian Zimmermann, Duygu Ceylan, Jimei Yang et al.

Using BERT for sequential sentence classification achieves state-of-the-art results on four datasets.

Arman Cohan, Iz Beltagy, Daniel King et al.

Transforming recommendation into a goal-oriented dialogue game, using self-supervised Bot-Play training, significantly improves interactive recommendation performance.

Dongyeop Kang, Anusha Balakrishnan, Pararth Shah et al.

Proposes an end-to-end multi-modality MOT framework (mmMOT) integrating deep features of image and point cloud, boosting robustness and accuracy.

Wenwei Zhang, Hui Zhou, Shuyang Sun et al.

Proposes an end-to-end radar odometry system using learned feature embeddings, reducing errors by 68% and running ten times faster than state-of-the-art.

Dan Barnes, Rob Weston, Ingmar Posner

Periodic driving trains disordered solids to achieve desired elastic properties.

Daniel Hexner, Andrea J. Liu, Sidney R. Nagel

Introduces control tasks and selectivity metric to evaluate probe interpretability; finds second-layer ELMo representations are more linguistically meaningful.

John Hewitt, Percy Liang

MoverScore combines contextual embeddings and Earth Mover's Distance for high semantic correlation evaluation.

Wei Zhao, Maxime Peyrard, Fei Liu et al.

Introduces TabFact dataset and combines Table-BERT with LPA models for table-based fact verification, achieving ~65% accuracy, below human performance.

Wenhu Chen, Hongmin Wang, Jianshu Chen et al.

Holistic++ framework uses MCMC to optimize single-view 3D scene parsing and human pose estimation, significantly improving performance.

Yixin Chen, Siyuan Huang, Tao Yuan et al.

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

Nir Levine, Yinlam Chow, Rui Shu et al.

Proposes regularization techniques (sample elimination, early stopping, max entropy adversarial RL) to improve transfer learning generalization in deep RL.

Suzan Ece Ada, Emre Ugur, H. Levent Akin

Oxford Radar RobotCar Dataset uses millimeter-wave FMCW radar for robust urban scene understanding, covering 280km in diverse weather conditions.

Dan Barnes, Matthew Gadd, Paul Murcutt et al.

This study systematically evaluates pretrained language models (like BERT) for factual knowledge storage without fine-tuning, showing strong performance in relation extraction and open-domain QA.

Fabio Petroni, Tim Rocktäschel, Patrick Lewis et al.

PlotQA with 28.9M QA pairs over 224k real-world plots advances complex reasoning in scientific visual question answering.

Nitesh Methani, Pritha Ganguly, Mitesh M. Khapra et al.

Using geometric analysis, this study reveals that BERT, ELMo, and GPT-2 produce highly anisotropic, layer-dependent, context-sensitive word embeddings with less than 5% variance explained by static vectors.

Kawin Ethayarajh

Mining commonsense knowledge using pretrained models without updating weights, outperforming supervised methods.

Joshua Feldman, Joe Davison, Alexander M. Rush