Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data.

Flamingo: a Visual Language Model for Few-Shot Learning

State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at this https URL.

Learning Transferable Visual Models From Natural Language Supervision

The identification of Hate Speech in Social Media is of great importance and receives much attention in the text classification community. There is a huge demand for research for languages other than English. The HASOC track intends to stimulate development in Hate Speech for Hindi, German and English. Three datasets were developed from Twitter and Facebook and made available. Binary classification and more fine-grained subclasses were offered in 3 subtasks. For all subtasks, 321 experiments were submitted. The approaches used most often were LSTM networks processing word embedding input. The performance of the best system for identification of Hate Speech for English, Hindi, and German was a Marco-F1 score of 0.78, 0.81 and 0.61, respectively.

/pdf/overview-of-the-hasoc-track-at-fire-2019-hate-speech-and-27y6nbsevs.pdf

Overview of the HASOC track at FIRE 2019: Hate Speech and Offensive Content Identification in Indo-European Languages

Data-driven and machine learning based approaches for detecting, categorising and measuring abusive content such as hate speech and harassment have gained traction due to their scalability, robustness and increasingly high performance. Making effective detection systems for abusive content relies on having the right training datasets, reflecting a widely accepted mantra in computer science: Garbage In, Garbage Out. However, creating training datasets which are large, varied, theoretically-informed and that minimize biases is difficult, laborious and requires deep expertise. This paper systematically reviews 63 publicly available training datasets which have been created to train abusive language classifiers. It also reports on creation of a dedicated website for cataloguing abusive language data hatespeechdata.com. We discuss the challenges and opportunities of open science in this field, and argue that although more dataset sharing would bring many benefits it also poses social and ethical risks which need careful consideration. Finally, we provide evidence-based recommendations for practitioners creating new abusive content training datasets.

/pdf/directions-in-abusive-language-training-data-a-systematic-1f88lrctry.pdf

Directions in abusive language training data, a systematic review: Garbage in, garbage out.

We present a novel task and dataset for evaluating the ability of vision and language models to conduct visio-linguistic compositional reasoning, which we call Winoground. Given two images and two captions, the goal is to match them correctly-but crucially, both captions contain a completely identical set of words, only in a different order. The dataset was carefully hand-curated by expert annotators and is labeled with a rich set offine-grained tags to assist in analyzing model performance. We probe a diverse range of state-of-the-art vision and language models and find that, surprisingly, none of them do much better than chance. Evidently, these models are not as skilled at visio-linguistic compositional reasoning as we might have hoped. We perform an extensive analysis to obtain insights into how future work might try to mitigate these models' shortcomings. We aim for Winoground to serve as a useful evaluation set for advancing the state of the art and driving further progress in the field. The dataset is available at https://huggingface.co/datasets/facebook/winoground.

Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality

This work proposes a new challenge set for multimodal classification, focusing on detecting hate speech in multimodal memes. It is constructed such that unimodal models struggle and only multimodal models can succeed: difficult examples ("benign confounders") are added to the dataset to make it hard to rely on unimodal signals. The task requires subtle reasoning, yet is straightforward to evaluate as a binary classification problem. We provide baseline performance numbers for unimodal models, as well as for multimodal models with various degrees of sophistication. We find that state-of-the-art methods perform poorly compared to humans (64.73% vs. 84.7% accuracy), illustrating the difficulty of the task and highlighting the challenge that this important problem poses to the community.

/pdf/the-hateful-memes-challenge-detecting-hate-speech-in-2jjy1293bv.pdf

The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes

/pdf/the-hateful-memes-challenge-competition-report-3s1ksboldx.pdf

Aravind Mohan

Papers

The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes

The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes

The Hateful Memes Challenge: Competition Report