Recent vision-language models outperform vision-only models on many image classification tasks. However, because of the absence of paired text/image descriptions, it remains difficult to fine-tune these models for fine-grained image classification. In this work, we propose a method, GIST, for generating image-specific fine-grained text descriptions from image-only datasets, and show that these text descriptions can be used to improve classification. Key parts of our method include 1. prompting a pretrained large language model with domain-specific prompts to generate diverse fine-grained text descriptions for each class and 2. using a pretrained vision-language model to match each image to label-preserving text descriptions that capture relevant visual features in the image. We demonstrate the utility of GIST by fine-tuning vision-language models on the image-and-generated-text pairs to learn an aligned vision-language representation space for improved classification. We evaluate our learned representation space in full-shot and few-shot scenarios across four diverse fine-grained classification datasets, each from a different domain. Our method achieves an average improvement of $4.1\%$ in accuracy over CLIP linear probes and an average of $1.1\%$ improvement in accuracy over the previous state-of-the-art image-text classification method on the full-shot datasets. Our method achieves similar improvements across few-shot regimes. Code is available at https://github.com/emu1729/GIST.

Various design settings for in-context learning (ICL), such as the choice and order of the in-context examples, can bias a model toward a particular prediction without being reflective of an understanding of the task. While many studies discuss these design choices, there have been few systematic investigations into categorizing them and mitigating their impact. In this work, we define a typology for three types of label biases in ICL for text classification: vanilla-label bias, context-label bias, and domain-label bias (which we conceptualize and detect for the first time). Our analysis demonstrates that prior label bias calibration methods fall short of addressing all three types of biases. Specifically, domain-label bias restricts LLMs to random-level performance on many tasks regardless of the choice of in-context examples. To mitigate the effect of these biases, we propose a simple bias calibration method that estimates a language model's label bias using random in-domain words from the task corpus. After controlling for this estimated bias when making predictions, our novel domain-context calibration significantly improves the ICL performance of GPT-J and GPT-3 on a wide range of tasks. The gain is substantial on tasks with large domain-label bias (up to 37% in Macro-F1). Furthermore, our results generalize to models with different scales, pretraining methods, and manually-designed task instructions, showing the prevalence of label biases in ICL.

The burgeoning progress in the field of Large Language Models (LLMs) heralds significant benefits due to their unparalleled capacities. However, it is critical to acknowledge the potential misuse of these models, which could give rise to a spectrum of social and ethical dilemmas. Despite numerous preceding efforts centered around distinguishing synthetic text, most existing detection systems fail to identify data synthesized by the latest LLMs, such as ChatGPT and GPT-4. In response to this challenge, we introduce an unpretentious yet potent detection approach proficient in identifying synthetic text across a wide array of fields. Moreover, our detector demonstrates outstanding performance uniformly across various model architectures and decoding strategies. It also possesses the capability to identify text generated utilizing a potent detection-evasion technique. Our comprehensive research underlines our commitment to boosting the robustness and efficiency of machine-generated text detection mechanisms, particularly in the context of swiftly progressing and increasingly adaptive AI technologies.

Large language models (LLMs) have recently soared in popularity due to their ease of access and the unprecedented ability to synthesize text responses to diverse user questions. However, LLMs like ChatGPT present significant limitations in supporting complex information tasks due to the insufficient affordances of the text-based medium and linear conversational structure. Through a formative study with ten participants, we found that LLM interfaces often present long-winded responses, making it difficult for people to quickly comprehend and interact flexibly with various pieces of information, particularly during more complex tasks. We present Graphologue, an interactive system that converts text-based responses from LLMs into graphical diagrams to facilitate information-seeking and question-answering tasks. Graphologue employs novel prompting strategies and interface designs to extract entities and relationships from LLM responses and constructs node-link diagrams in real-time. Further, users can interact with the diagrams to flexibly adjust the graphical presentation and to submit context-specific prompts to obtain more information. Utilizing diagrams, Graphologue enables graphical, non-linear dialogues between humans and LLMs, facilitating information exploration, organization, and comprehension.

Multimodal learning, which aims to understand and analyze information from multiple modalities, has achieved substantial progress in the supervised regime in recent years. However, the heavy dependence on data paired with expensive human annotations impedes scaling up models. Meanwhile, given the availability of large-scale unannotated data in the wild, self-supervised learning has become an attractive strategy to alleviate the annotation bottleneck. Building on these two directions, self-supervised multimodal learning (SSML) provides ways to learn from raw multimodal data. In this survey, we provide a comprehensive review of the state-of-the-art in SSML, in which we elucidate three major challenges intrinsic to self-supervised learning with multimodal data: (1) learning representations from multimodal data without labels, (2) fusion of different modalities, and (3) learning with unaligned data. We then detail existing solutions to these challenges. Specifically, we consider (1) objectives for learning from multimodal unlabeled data via self-supervision, (2) model architectures from the perspective of different multimodal fusion strategies, and (3) pair-free learning strategies for coarse-grained and fine-grained alignment. We also review real-world applications of SSML algorithms in diverse fields such as healthcare, remote sensing, and machine translation. Finally, we discuss challenges and future directions for SSML. A collection of related resources can be found at: https://github.com/ys-zong/awesome-self-supervised-multimodal-learning.

Electronic health records (EHRs) store an extensive array of patient information, encompassing medical histories, diagnoses, treatments, and test outcomes. These records are crucial for enabling healthcare providers to make well-informed decisions regarding patient care. Summarizing clinical notes further assists healthcare professionals in pinpointing potential health risks and making better-informed decisions. This process contributes to reducing errors and enhancing patient outcomes by ensuring providers have access to the most pertinent and current patient data. Recent research has shown that incorporating prompts with large language models (LLMs) substantially boosts the efficacy of summarization tasks. However, we show that this approach also leads to increased output variance, resulting in notably divergent outputs even when prompts share similar meanings. To tackle this challenge, we introduce a model-agnostic Soft Prompt-Based Calibration (SPeC) pipeline that employs soft prompts to diminish variance while preserving the advantages of prompt-based summarization. Experimental findings on multiple clinical note tasks and LLMs indicate that our method not only bolsters performance but also effectively curbs variance for various LLMs, providing a more uniform and dependable solution for summarizing vital medical information.

Over the past few months, shiny metallic orbs have materialized cities around the world, from New York to Berlin to Tokyo. Its detractors slam them as invasive, dystopian and exploitative. Welcome to the rollout of Worldcoin, an AI-meets-crypto project from OpenAI founder Sam Altman that has stirred endless controversy. The startup uses orbs to scan people's eyes in exchange for a digital ID and possibly some cryptocurrency, depending on what country they live in. Altman and his co-founder Alex Blania hope that Worldcoin will provide a new solution to online identity in a digital landscape rife with scams, bots and even AI imposters.

A larger-than-life Michelle Donelan beams on to a screen in Google's London headquarters. The UK science and innovation secretary is appearing via video to praise the US tech behemoth for its plans to equip workers and bosses with basic skills in artificial intelligence (AI). "The recent explosion in the use of AI tools like ChatGPT and Google's Bard show that we are on the cusp of a new and exciting era in artificial intelligence, and it is one that will dramatically improve people's lives," says Donelan. Google's "ambitious" training programme is "so important" and "exceptional in its breadth", she gushes in a five-minute video, filmed in her ministerial office. Welcome to the AI arms race, where nations are bending over backwards to attract cash and research into the nascent technology.

Emergent chain-of-thought (CoT) reasoning capabilities promise to improve performance and explainability of large language models (LLMs). However, uncertainties remain about how reasoning strategies formulated for previous model generations generalize to new model generations and different datasets. In this small-scale study, we compare different reasoning strategies induced by zero-shot prompting across six recently released LLMs (davinci-002, davinci-003, GPT-3.5-turbo, GPT-4, Flan-T5-xxl and Cohere command-xlarge) on a mixture of six question-answering datasets, including datasets from scientific and medical domains. Our findings demonstrate that while some variations in effectiveness occur, gains from CoT reasoning strategies remain robust across different models and datasets. GPT-4 has the most benefit from current state-of-the-art reasoning strategies and exhibits the best performance by applying a prompt previously discovered through automated discovery.

This work was supported in part by the Complex Human Data Hub at the University of Melbourne and by ARC FT190100200. Correspondence concerning this article should be addressed to Jerome Han. Abstract The impressive recent performance of large language models has led many to wonder to what extent they can serve as models of general intelligence or are similar to human cognition. We address this issue by applying GPT-3.5 and GPT-4 to a classic problem in human inductive reasoning known as property induction. Although GPT-3.5 struggles to capture many aspects of human behaviour, GPT-4 is much more successful: for the most part, its performance qualitatively matches that of humans, and the only notable exception is its failure to capture the phenomenon of premise non-monotonicity. Our work demonstrates that property induction allows for interesting comparisons between human and machine intelligence and provides two large datasets that can serve as benchmarks for future work in this vein.