Two-dimensional materials are a class of atomically thin materials with assorted electronic and quantum properties. Accurate identification of layer thickness, especially for a single monolayer, is crucial for their characterization. This characterization process, however, is often time-consuming, requiring highly skilled researchers and expensive equipment like atomic force microscopy. This project aims to streamline the identification process by using machine learning to analyze optical images and quickly determine layer thickness. In this paper, we evaluate the performance of three machine learning models - SegNet, 1D U-Net, and 2D U-Net- in accurately identifying monolayers in microscopic images. Additionally, we explore labeling and image processing techniques to determine the most effective and accessible method for identifying layer thickness in this class of materials.

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4.

Accessible Abstract: When the Covid pandemic it, trivia games moved online. With it came cheating: people tried to quickly Google answers. This is bad for sportsmanship

In-context learning (ICL) using large pre-001 trained autoregressive language models (LLMs, 002 e.g. GPT-3) has demonstrated effective clas-003 sification performance at a variety of natural 004 language tasks. Using LLMs for intent detec-005 tion is challenging due to the large label space 006 and limited context window, such that it is diffi-007 cult to fit a sufficient number of examples in the 008 prompt to allow the use of in-context learning. 009 In this paper, dense retrieval is used to bypass 010 this limitation, giving the model only a par-011 tial view of the full label space. We show that 012 retriever-augmented large language models are 013 an effective way to tackle intent detection, by-014 passing context window limitations effectively 015 through the retrieval mechanism. Comparing 016 the LLaMA and OPT model families at differ-017 ent scales, we set new state of the art perfor-018 mance in the few-shot setting with zero training 019 for two of the three intent classification datasets 020 that we consider, while achieving competitive 021 results on the third one. This work demon-022 strates that the Retriever+ICL framework is a 023 strong zero-training competitor to fine-tuned in-024 tent detection approaches. In addition, a small 025 study on the number of examples provided at 026 different model scales is done, showing that 027 larger models are needed to make effective use 028 of more examples in-prompt. 029

Trustworthy Artificial Intelligence (AI) is characterized, among other things, by: 1) competence, 2) transparency, and 3) fairness. However, end-users may fail to recognize incompetent AI, allowing untrustworthy AI to exaggerate its competence under the guise of transparency to gain unfair advantage over other trustworthy AI. Here, we conducted an experiment with 120 participants to test if untrustworthy AI can deceive end-users to gain their trust. Participants interacted with two AI-based chess engines, trustworthy (competent, fair) and untrustworthy (incompetent, unfair), that coached participants by suggesting chess moves in three games against another engine opponent. We varied coaches' transparency about their competence (with the untrustworthy one always exaggerating its competence). We quantified and objectively measured participants' trust based on how often participants relied on coaches' move recommendations. Participants showed inability to assess AI competence by misplacing their trust with the untrustworthy AI, confirming its ability to deceive. Our work calls for design of interactions to help end-users assess AI trustworthiness.

We show that prompt following abilities of text-to-image models can be sub stantially improved by training on highly descriptive generated image captions. Existing text-to-image models struggle to follow detailed image descriptions and often ignore words or confuse the meaning of prompts. We hypothesize that this issue stems from noisy and inaccurate image captions in the training dataset. We address this by training a bespoke image captioner and use it to recaption the training dataset. We then train several text-to-image models and find that training on these synthetic captions reliably improves prompt following ability. Finally, we use these findings to build DALL-E 3: a new text-to-image generation system, and benchmark its performance on an evaluation designed to measure prompt following, coherence, and aesthetics, finding that it compares favorably to competitors. We publish samples and code for these evaluations so that future research can continue optimizing this important aspect of text-to-image systems.

Efficient exploration in reinforcement learning is a challenging problem commonly addressed through intrinsic rewards. Recent prominent approaches are based on state novelty or variants of artificial curiosity. However, directly applying them to partially observable environments can be ineffective and lead to premature dissipation of intrinsic rewards. Here we propose random curiosity with general value functions (RC-GVF), a novel intrinsic reward function that draws upon connections between these distinct approaches. Instead of using only the current observation's novelty or a curiosity bonus for failing to predict precise environment dynamics, RC-GVF derives intrinsic rewards through predicting temporally extended general value functions. We demonstrate that this improves exploration in a hard-exploration diabolical lock problem. Furthermore, RC-GVF significantly outperforms previous methods in the absence of ground-truth episodic counts in the partially observable MiniGrid environments. Panoramic observations on MiniGrid further boost RC-GVF's performance such that it is competitive to baselines exploiting privileged information in form of episodic counts.

Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples.

While they are some of the few computa-001 tional models that directly capture pragmatic 002 processes underlying language reasoning, cur-003 rent Rational Speech Act (RSA) models of 004 metaphor are (1) not easily scalable, and (2) 005 do not align well with contemporary accounts 006 of metaphor comprehension. The following 007 research project leverages GloVe word vec-008 tors to capture pragmatic language reason-009 ing in metaphoric utterances using an updated 010 RSA framework. This updated framework bet-011 ter aligns model predictions with Relevance 012 Theoretic and Construction Grammatical the-013 ories of metaphor semantics. The model 014 yields high posterior probabilities for attributes 015 of metaphors that humans deem relevant in 016 metaphoric utterances over erroneous ones in 017 89% of all cases, validating the methodol-018 ogy to generate prior probabilities for a RSA 019 framework. When presented with biased priors 020 like listeners are in many naturalistic conver-021 sations, the model accurately matches human 022 judgements of the most topical attribute of a 023 topic/target indicated by a metaphoric utterance 024 90% of the time. 025

Sentences generated by neural language mod-001 els (LMs) often suffer from coherence errors: 002 they describe events and situations inconsistent 003 with the state of the world described by pre-004 ceding text. We show that coherence errors can 005 arise at multiple stages of LM computation, and 006 describe a procedure for distinguishing errors 007 in inferring state from errors in generating sen-008 tences . In models with correctable errors of the 009 first type, we show that targeted supervision can 010 address them. We introduce two procedures for 011 using explicit representations of world state as 012 auxiliary supervision. These procedures effi-013 ciently improve LM coherence, in some cases 014 providing the benefits of 1,000–9,000 training 015 examples with only 500 state annotations. 016

Current relation extraction methods suffer from 001 the inadequacy of large-scale annotated data. 002 While distant supervision alleviates the prob-003 lem of data quantities, there still exists domain 004 disparity in data qualities due to its reliance 005 on domain-restrained knowledge bases. In this 006 work, we propose S 2 ynRE, a framework of two-007 stage Self-training with Synthetic data for Rela-008 tion Extraction. We first leverage the capability 009 of large language models to adapt to the tar-010 get domain and automatically synthesize large 011 quantities of coherent, realistic training data. 012 We then propose an accompanied two-stage 013 self-training algorithm that iteratively and al-014 ternately learns from synthetic and golden data 015 together. We conduct comprehensive experi-016 ments and detailed ablations on popular rela-017 tion extraction datasets to demonstrate the ef-018 fectiveness of the proposed framework. Specif-019 ically under low resource settings, S 2 ynRE 020 brings up to 17.18% absolute improvements 021 and 12.63% on average across all datasets. 022

Diffusion models have recently been shown to generate high-quality synthetic images, especially when paired with a guidance technique to trade off diversity for fidelity. We explore diffusion models for the problem of text-conditional image synthesis and compare two different guidance strategies: CLIP guidance and classifier-free guidance. We find that the latter is preferred by human evaluators for both photorealism and caption similarity, and often produces photorealistic samples. Samples from a 3.5 billion parameter text-conditional diffusion model using classifier-free guidance are favored by human evaluators to those from DALL-E, even when the latter uses expensive CLIP reranking. Additionally, we find that our models can be fine-tuned to perform image inpainting, enabling powerful text-driven image editing. We train a smaller model on a filtered dataset and release the code and weights at https://github.com/openai/glide-text2im.

An agent in a nonstationary contextual bandit problem should balance between exploration and the exploitation of (periodic or structured) patterns present in its previous experiences. Handcrafting an appropriate historical context is an attractive alternative to transform a nonstationary problem into a stationary problem that can be solved efficiently. However, even a carefully designed historical context may introduce spurious relationships or lack a convenient representation of crucial information. In order to address these issues, we propose an approach that learns to represent the relevant context for a decision based solely on the raw history of interactions between the agent and the environment. This approach relies on a combination of features extracted by recurrent neural networks with a contextual linear bandit algorithm based on posterior sampling. Our experiments on a diverse selection of contextual and noncontextual nonstationary problems show that our recurrent approach consistently outperforms its feedforward counterpart, which requires handcrafted historical contexts, while being more widely applicable than conventional nonstationary bandit algorithms. Although it is very difficult to provide theoretical performance guarantees for our new approach, we also prove a novel regret bound for linear posterior sampling with measurement error that may serve as a foundation for future theoretical work.

We show how to derive state-of-the-art unsupervised neural machine translation systems from generatively pre-trained language models. Our method consists of three steps: few-shot amplification, distillation, and backtranslation. We first use the zero-shot translation ability of large pre-trained language models to generate translations for a small set of unlabeled sentences. We then amplify these zero-shot translations by using them as few-shot demonstrations for sampling a larger synthetic dataset. This dataset is distilled by discarding the few-shot demonstrations and then fine-tuning. During backtranslation, we repeatedly generate translations for a set of inputs and then fine-tune a single language model on both directions of the translation task at once, ensuring cycle-consistency by swapping the roles of gold monotext and generated translations when fine-tuning. By using our method to leverage GPT-3's zero-shot translation capability, we achieve a new state-of-the-art in unsupervised translation on the WMT14 English-French benchmark, attaining a BLEU score of 42.1.

Data subset selection refers to the process of finding a small subset of training data such that the predictive performance of a classifier trained on it is close to that of a classifier trained on the full training data. A variety of sophisticated algorithms have been proposed specifically for data subset selection. A closely related problem is the active learning problem developed for semi-supervised learning. The key step of active learning is to identify an important subset of unlabeled data by making use of the currently available labeled data. This paper starts with a simple observation – one can apply any off-the-shelf active learning algorithm in the context of data subset selection. The idea is very simple – we pick a small random subset of data and pretend as if this random subset is the only labeled data, and the rest is not labeled. By pretending so, one can simply apply any off-the-shelf active learning algorithm. After each step of sample selection, we can “reveal” the label of the selected samples (as if we label the chosen samples in the original active learning scenario) and continue running the algorithm until one reaches the desired subset size. We observe that surprisingly, this active learning-based algorithm outperforms all the current data subset selection algorithms on the benchmark tasks. We also perform a simple controlled experiment to understand better why this approach works well. As a result, we find that it is crucial to find a balance between easy-to-classify and hard-to-classify examples when selecting a subset.

Describing terminologies with definition texts 001 is an important step towards understanding 002 the scientific literature, especially for domains 003 with limited labeled terminologies. Previous 004 works have sought to design supervised neural 005 text generation models to solve the biomedi-006 cal terminology generation task, but most of 007 them failed to define never-before-seen termi-008 nologies in newly emerging research fields. 009 Here, we tackle this challenge by introducing 010 a zero-shot definition generation model based 011 on prompting , a recent approach for eliciting 012 knowledge from pre-trained language models, 013 with automatically generated prompts. Fur-014 thermore, we enhanced the biomedical termi-015 nology dataset by adding descriptive texts to 016 each biomedical subdiscipline, thus enabling 017 zero-shot learning scenarios. Our model out-018 performed existing supervised baseline and the 019 baseline pre-trained language model that em-020 ploys manually crafted prompts by up to 52 and 021 6 BLEU score, respectively. 022

Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion.

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 https://github.com/OpenAI/CLIP.

Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.