Arts Benchmark Proposal 
Google Research AI blog
JUNE 2, 2023
The resulting AVFormer model achieves state-of-the-art zero-shot performance on three different AV-ASR benchmarks (How2, VisSpeech and Ego4D ), while also crucially preserving decent performance on traditional audio-only speech recognition benchmarks (i.e., LibriSpeech ). Unconstrained audiovisual speech recognition.
Google Research AI blog
FEBRUARY 8, 2023
Using data-centric approaches, we show state-of-the-art results in both. In “ SPADE: Semi-supervised Anomaly Detection under Distribution Mismatch ”, we propose a novel semi-supervised AD framework that yields robust performance even under distribution mismatch with limited labeled samples. We consider methods with both shallow (e.g.,
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Google Research AI blog
MAY 26, 2023
In light of these challenges, we propose “ MatCha: Enhancing Visual Language Pretraining with Math Reasoning and Chart Derendering ”. We also propose “ DePlot: One-shot visual language reasoning by plot-to-table translation ”, a model built on top of MatCha for one-shot reasoning on charts via translation to tables.
Google Research AI blog
AUGUST 28, 2023
Various techniques such as image-text pre-training , knowledge distillation , pseudo labeling , and frozen models, often employing convolutional neural network (CNN) backbones, have been proposed. To address this, we propose cropped positional embeddings (CPE). We are also releasing the code here.
Google Research AI blog
FEBRUARY 7, 2023
We proposed a theoretical architecture that would “remember events” in the form of sketches stored in an external LSH table with pointers to modules that process such sketches. We have proposed a new constrained optimization algorithm for automating hyperparameter tuning. T5-Large models have <1% nonzero entries.
Google Research AI blog
APRIL 20, 2023
A number of neural network–based solutions have been able to show good performance on benchmarks and also support the above criterion. However, other work has suggested that even linear models can outperform these transformer variants on time-series benchmarks. Left: MSE on the test set of a popular traffic forecasting benchmark.
Google Research AI blog
MAY 18, 2023
We demonstrate that this model is scalable, can be adapted to large pre-trained ViTs without requiring full fine-tuning, and achieves state-of-the-art results across many video classification benchmarks. Importantly, it outperforms all state-of-the-art methods trained jointly on image+video datasets.
Expert insights. Personalized for you.
103 
Let's personalize your content