AdaRound uses Taylor expansion-based adaptive rounding, outperforming nearest neighbor, achieving <1% accuracy loss on 4-bit ResNet without fine-tuning.

Markus Nagel, Rana Ali Amjad, Mart van Baalen et al.

Models are optimized via multi-objective framework; nucleus sampling outperforms others; likelihood trap identified and addressed.

Hugh Zhang, Daniel Duckworth, Daphne Ippolito et al.

Proposes logical NLG from open-domain tables, using Transformer and pre-trained models, achieving significant gains in logical fidelity.

Wenhu Chen, Jianshu Chen, Yu Su et al.

This paper analyzes global linguistic resource distribution, proposes a six-category classification, highlights disparities affecting multilingual NLP, and urges community focus on low-resource languages.

Pratik Joshi, Sebastin Santy, Amar Budhiraja et al.

Proposes a hybrid CEM and gradient descent method to improve high-dimensional model predictive control efficiency.

Homanga Bharadhwaj, Kevin Xie, Florian Shkurti

Shortcut learning in deep learning affects model generalization in complex scenarios.

Robert Geirhos, Jörn-Henrik Jacobsen, Claudio Michaelis et al.

DISTS combines structure and texture similarity, improving robustness to texture resampling and geometric transformations in image quality assessment.

Keyan Ding, Kede Ma, Shiqi Wang et al.

Using Random Weight Guessing (RWG), the study evaluates RL benchmarks by analyzing score distributions of randomly initialized neural networks, revealing environment complexity.

Declan Oller, Tobias Glasmachers, Giuseppe Cuccu

INLP (Iterative Nullspace Projection) removes bias from neural representations by iterative training of linear classifiers and nullspace projection, significantly reducing gender bias in word embeddings.

Shauli Ravfogel, Yanai Elazar, Hila Gonen et al.

HybridQA integrates heterogeneous data with a multi-hop reasoning framework using SQLNet and BERT, achieving over 52% EM on complex questions.

Wenhu Chen, Hanwen Zha, Zhiyu Chen et al.

TVOS propagates labels through long-range spatiotemporal embeddings, reaching 72.3 J&F on DAVIS 2017 validation at 37 FPS.

Yizhuo Zhang, Zhirong Wu, Houwen Peng et al.

Using spectral analysis of Schrödinger operators, the paper reveals that SGD's linear convergence rate depends exponentially on the learning rate, especially in nonconvex optimization.

Bin Shi, Weijie J. Su, Michael I. Jordan

Pretrained Transformers like BERT show improved out-of-distribution robustness across seven NLP datasets.

Dan Hendrycks, Xiaoyuan Liu, Eric Wallace et al.

The paper proposes CT, a three-sample nonparametric Mann–Whitney test for detecting generative-model data-copying.

Casey Meehan, Kamalika Chaudhuri, Sanjoy Dasgupta

Proposes FLOPs-regularized high-dimensional sparse embeddings, achieving 10x speedup with comparable accuracy.

Biswajit Paria, Chih-Kuan Yeh, Ian E. H. Yen et al.

NNV employs set-based algorithms (polyhedra, Star, Zonotope, ImageStar) for verifying DNNs and learning-enabled CPS, achieving significant speedups over existing tools.

Hoang-Dung Tran, Xiaodong Yang, Diego Manzanas Lopez et al.

Introduces Longformer with linear attention combining local windows and global tokens, enabling efficient long document modeling surpassing RoBERTa.

Iz Beltagy, Matthew E. Peters, Arman Cohan

BLEURT, based on BERT with synthetic pretraining, significantly improves text generation evaluation accuracy and robustness.

Thibault Sellam, Dipanjan Das, Ankur P. Parikh

Proposes an improved cutting plane algorithm with O(n log(κ)) oracle evaluations and O(n^2) per evaluation, optimizing convex and convex-concave problems.

Haotian Jiang, Yin Tat Lee, Zhao Song et al.

QAGS uses question answering to evaluate factual consistency of summaries, outperforming ROUGE with a correlation of 54.52 on CNN/DM.

Alex Wang, Kyunghyun Cho, Mike Lewis